2.0.0-beta.6

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

CHANGELOG.md

@@ -4,6 +4,93 @@ All notable changes to LLeMbas CLI. The format follows
 [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project uses
 [Semantic Versioning](https://semver.org/).
 
+## [2.0.0-beta.6] — 2026-11-10
+
+Harness spec 3.4.0 (spec commits 4b8786b, b66891e): one progress vocabulary for every client, and
+subagents organised as the main agent's tree.
+
+### Added
+
+- **Progress phases.** The line under the conversation says what the agent is doing in the words
+  every LLeMbas client uses — *Queued*, *Waiting for {model}*, *Reading the prompt: n / total
+  tokens (p%)*, *Thinking*, *Writing: n tok/s*, *Running {tool}: {what}*, *Waiting for you*,
+  *Waiting for subagents*, *Compacting the conversation*, *Retrying in n s*, *Stopped*, *Failed* —
+  with how long it has lasted, then the turn's time and tokens. The CLI keeps its own estimate of
+  how long the rest of a long prompt will take. Over the link each phase is a `phase` event and is
+  in the session's status (capability `progress`).
+- **The main agent and its subagents.** Only the main agent adds, reassigns (`team_assign`),
+  replaces (`team_replace`), stops and decides; a subagent added with `lead: true` is a
+  sub-manager that directs the subagents placed under it (`parent`) and adds none. Each subagent
+  has an id (`s1`, `s2`, …), a name, a position, a parent, a task, a mode and a model, and can ask
+  `team_whoami`. With Min above 1 the main agent starts with Min − 1 open slots to fill.
+- **Permissions only go down.** A subagent's mode is its parent's or stricter. A narrowed one may
+  be given request options (read, write, execute) and ask the main agent with `team_request`; the
+  main agent decides with `team_decide`, and never allows more than its own mode would. What
+  nobody in the tree could allow — a `bash` call in an all-Edit tree — is put to you, on an
+  approval card that names the subagent.
+- **A subagent's end.** It reports and ends with `team_finish`. Its findings (`team_findings`) are
+  kept after it ends. When its context fills it does not compact: it is paused, gets one last
+  turn of at most 8 steps to save its findings and tell its parent, and ends; the main agent
+  replaces it or uses what it found.
+- **Seeing the tree.** `/team` draws every agent with its position, state, model, mode, context
+  use, current phase and messages, the open slots and the subagents that ended; `/team open
+  <name>` shows a subagent's session read-only, live or after it ended; `/team findings <name>`;
+  `/context <name>` for a subagent's context.
+- **An agent chat's subagents run on this device.** The chat's Min and Max come with every turn
+  (capability `agents`) and the tree, progress, findings and transcripts go back to the server
+  (capability `tree`: the `team` event with the tree, `_lembas/team`, `/transcript`, `/findings`,
+  `/stop`).
+
+### Changed
+
+- A subagent can no longer add subagents of its own or start a `task`; it asks the main agent.
+  `team_stop` is the main agent's alone.
+
+## [2.0.0-beta.5] — 2026-11-09
+
+Harness spec 3.3.0 (spec commit 02c7282): subagents run on your model unless you name another,
+and `/context` shows where the window goes the way the side panel of the other apps does.
+
+### Changed
+
+- **Subagents run on the session's model**, unless you named another model in this session's own
+  messages — its id, the part after its provider, or its display name, as a whole word, in any
+  letter case. A `task` or `team_add` that asks for any other model is refused ("Use another
+  model for subagents only when the person names it."), and an agent file's own `model:` is no
+  longer used. Nothing in the settings chooses a subagent's model.
+- **Only the logical capacity limits remain.** A named model on the same one-model server as the
+  session's model is refused ("That model is on the same one-model server as this chat's model;
+  name a model on another server."). On a model that serves one request at a time, a subagent on
+  that same model is refused ("This model serves one request at a time: name another model, on
+  another server, for subagents.") and nothing else is picked instead.
+- **Subagents are offered** — the size setting, the team tools — when subagents are on and some
+  model could serve one: the session's, when it serves two requests at once, or any model of the
+  server that is not on its one-model server. When the session's model cannot serve its own
+  subagents, the agent is told so in its first note, and `/effort`, `/team size` and `/team` say it
+  beside the setting.
+- **"Subagents" is the word** for the setting, with **Min** and **Max**: the row beside the effort
+  in `/effort`, `/team size`, `/settings`. "Helper" is gone. The status bar reads
+  `agents 3/4 · 2 subagents working` while there is a team.
+- **The numbers count this agent**: `Subagents: Min 1, Max 4 — 1 is this agent alone`. Max 1 is
+  the agent alone, with no subagents; Max 4 is the agent and up to three subagents.
+- **The setting is one range with two handles.** In `/effort` and `/team size` the Subagents row
+  shows `Min 1`, a track from 1 to the cap with a handle for each, and `Max 4`: ←→ move the handle in focus, tab switches to the other, enter
+  sets both for the session (Min never passes Max; Max goes up to `agents.cap`). `/team size` also
+  keeps the usual pairs and `<min> <max>` typed.
+- **`/context`** shows the window as used tokens of the window and a percentage, one bar split by
+  category, every category with its tokens and its share of the window, what is free, and how far
+  it is to automatic compaction (`N until auto-compact`, with `/compact` to do it now).
+
+## [2.0.0-beta.4] — 2026-11-09
+
+### Fixed
+
+- **`lembas login` no longer loops on a sign-in the server stopped accepting.** With an earlier login
+  for the same server kept here (from an older install, or LLeMbas Desktop), login only refreshed it;
+  when the server refused the old token it said “run login again”, and running it again did the same.
+  Now a refused sign-in starts a new one, which replaces the old.
+- The installer's last line says to run `lembas login`, not to edit `connections.yaml` (gone in 2.0).
+
 ## [2.0.0-beta.3] — 2026-11-09
 
 Harness spec 3.2.0: teams, effort levels that mean what they say, the model's reasoning sent

harness/README.md

@@ -15,17 +15,19 @@ page *Harness-parity* in LLeMbas CLI's wiki.
 |---|---|
 | `VERSION` | The spec's own version, apart from LLeMbas CLI's |
 | `purpose.json` | The `purpose` argument added to a tool whose calls say what they are for |
-| `tools/<name>.json` | One tool each: name, `scope`, `sessions`, `risk`, `exclusive`, `block` (how a call is shown: `shell`, `edit`, `read`, `search`, `web`, `task`, `plan` — not drawn, the plan card refreshes — `hidden`, `tool`), `purpose`, former names (`aliases`, per project), renamed arguments, `variables`, what a project offers of the parameters (`narrow`, per project), `offered` (`subagent`: given to subagents only, never to the main agent — `report`), `requires` and `unavailable` (when it can be offered, and what a call answers when that has gone), `confirm` (`always`: the person is asked in every mode), the description and the JSON Schema of its parameters — exactly what a model is sent |
+| `tools/<name>.json` | One tool each: name, `scope`, `sessions`, `risk`, `exclusive`, `block` (how a call is shown: `shell`, `edit`, `read`, `search`, `web`, `task`, `plan` — not drawn, the plan card refreshes — `hidden`, `tool`), `purpose`, former names (`aliases`, per project), renamed arguments, `variables`, what a project offers of the parameters (`narrow`, per project), `offered` (`subagent`: given to subagents only, never to the main agent — `report`, `team_finish`, `team_request`; `main`, spec 3.4.0: given to the main agent only, never to a subagent — `team_add`, `team_assign`, `team_replace`, `team_stop`, `team_decide`), `requires` and `unavailable` (when it can be offered, and what a call answers when that has gone), `confirm` (`always`: the person is asked in every mode), the description and the JSON Schema of its parameters — exactly what a model is sent |
 | `prompts/` | The prompt texts; `prompts/index.json` says which are shared, which LLeMbas fragment each one is, and what each `{{variable}}` in a text means. `prompts/personality/` holds the personality presets both offer |
 | `permission/hardline.json` | The floor: commands refused in every mode, protected paths, and how a command is spelled plainly before it is checked again |
 | `permission/arity.json` | How much of a command an "always allow" covers |
 | `permission/defaults.json` | The rules every session starts from |
 | `permission/modes.json` | The four modes — manual, edit, auto, plan — as what each does with each risk class |
 | `models/schema.json` | Model metadata: context, output limit, efforts and the thinking switch, vision, tools, notes, capacity (`parallel`); `levels` — the effort levels a chat offers and what each sends |
-| `loop.json` | How one prompt runs: the step backstop, budgets, wrap-up, nudges, parallel calls, capacity; `jobs` — background work, its marks and limits; `team` — a team's size, names, marks, refusals and texts; `context` — result ids, stubs, the output cap, the compaction table, the environment lines, the session notes |
+| `loop.json` | How one prompt runs: the step backstop, budgets, wrap-up, nudges, parallel calls, capacity; `jobs` — background work, its marks and limits; `subagents` — the model refusals, what names a model, the setting's label; `team` — a team's size, names, marks, refusals and texts; `progress` — the phases' texts; `context` — result ids, stubs, the output cap, the compaction table, the environment lines, the session notes |
 | `filesystem.md` | The file system a session works in: paths, node kinds, the four access levels, errors, listings, search and indexing by tokens, stacked instruction files, skills, memory |
 | `context.md` | Context hygiene: what happens to a tool result when it is inserted, and never afterwards; the prefix that stays still; reasoning sent back during its turn; no absolute paths in the prompt |
 | `team.md` | Teams: a lead and the members it adds, the size setting, the room, who receives what, reports, a manager's refusals, capacity, stopping |
+| `subagents.md` | Subagents: the word, which model one runs on (the session's, or one the person named), the capacity refusals, and when the Subagents setting is offered; from 3.4.0 (§5–§13) the main agent and its tree — only it adds, permissions inherited and never widened, request options, a subagent's life and its findings, what clients show |
+| `progress.md` | Progress (3.4.0): the thirteen phases every client shows the same way, the data each carries, their texts, and where they are sent |
 | `quirks.json` | What real servers do, and how both harnesses cope |
 | `conformance/<area>.json` | Cases with their expected results, run by both test suites |
 | `acp.md` | The link between them: ACP plus every `_lembas/*` method, field and capability |
@@ -33,7 +35,7 @@ page *Harness-parity* in LLeMbas CLI's wiki.
 ## Vocabulary
 
 - **Risk classes:** `read`, `write`, `execute`, `interact`. **Reading and writing are never one
-  tool** among the shared ones, so a read-only helper is given exactly the readers.
+  tool** among the shared ones, so a read-only subagent is given exactly the readers.
 - **Modes:** `manual`, `edit`, `auto`, `plan` (`unrestricted` is read as `auto`).
 - **Scopes:** `shared` — both projects implement it; `cli` or `llembas` — only that one. The same
   two names key what is per project elsewhere: a tool's `aliases`, `loop.json`'s budgets, a
@@ -48,8 +50,9 @@ page *Harness-parity* in LLeMbas CLI's wiki.
   page of the chat is open and attached (not a schedule, not a background job); `display`: the
   session is shown somewhere a file can be opened (the web UI, Desktop; not a bare terminal);
   `forge`: the session's view holds a repository mounted from a forge; `plan`: the session is in
-  plan mode; `team` (3.2.0): the session's team setting allows more than one agent, on a model that
+  plan mode; `team` (3.2.0): the session's team setting allows more than one agent, and a model could
-  serves two requests at once (`team.md` §2). A requirement that goes away between the offer and the call makes the call answer the
+  serve as a subagent (3.3.0: `subagents.md` §4; before, the chat's model serving two requests at
+  once). A requirement that goes away between the offer and the call makes the call answer the
   tool's `unavailable` text, word for word, as its result. **From 3.1.0 a requirement that can come
   and go during a session does not take the tool out of the list** (`context.md` §10, rule 4): the
   tool is offered from the start and answers `unavailable` while the requirement is not met.
@@ -68,6 +71,68 @@ page *Harness-parity* in LLeMbas CLI's wiki.
   approval in plan mode and answers its `unavailable` text outside it. Their permission key is
   `plan` (`permission/defaults.json`), which a rule written for the former `todo` key still reaches.
 
+## 3.4.0
+
+Written ahead of both implementations; the server vendors it as written, both carry it in the same
+round (`progress.md`, `subagents.md` §5–§13).
+
+- **Progress** (`progress.md`, `loop.json` `progress`): one fixed set of phases — `queued`,
+  `waiting_for_model`, `loading_model`, `reading_prompt`, `thinking`, `writing`, `running_tool`,
+  `waiting_for_person`, `waiting_for_subagents`, `compacting`, `retrying`, `stopped`, `failed` —
+  each a structured object `{ phase, since, …data }` with fixed fields, an English text that is
+  the shared i18n key, and one line every client draws: `text · elapsed`. Sent beside the old
+  `status` text in `/api/v2` and over the link (`acp.md` capability `progress`: the `phase` event,
+  `progress` in `status`); `status` stays. Area `progress`.
+- **Subagents v2** (`subagents.md` §5–§13; supersedes `team.md` where they differ, which says so):
+  the **main agent** is the session's, permanent, and **alone adds, assigns, replaces, stops and
+  decides** (`offered: main`); a member with `lead: true` is a **sub-manager** that directs those
+  placed under it (`team_add`'s `parent`). Every subagent has an id (`s1`…), a name, a position, a
+  parent, a task, a mode and a model. Min above 1 opens `min − 1` **slots** from the start of the
+  reply. **Permissions are inherited and never widened**: a subagent's mode is its parent's or
+  stricter; a narrowed one may get **request options** (`read`, `write`, `execute`) and ask the
+  main agent (`team_request` → `team_decide`, a grant that never reaches past the main agent's own
+  mode); a call nobody in the tree can allow goes **to the person**. A subagent ends with
+  `team_finish`, is stopped, or **runs out of context**: paused, one final turn to save its
+  findings and tell its parent, then `out_of_context` and a report; findings and transcripts are
+  kept, readable by the parent, the main agent and the person. New tools `team_whoami`,
+  `team_finish`, `team_findings`, `team_request`, `team_decide`, `team_assign`, `team_replace`;
+  `team_add` gains `parent`, `mode`, `request`; `team_send` takes `parent`; `team_stop` is the main
+  agent's. `blocks/team_submanager.md`; `team_lead.md` and `team_member.md` reworded. Over the link,
+  capability `tree`: the `team` event, `_lembas/team`, `/transcript` (read-only), `/findings`,
+  `/stop`, `subagent` on an approval. **An agent chat's subagents are the device's** (`subagents.md`
+  §14, `acp.md` *Subagents in an agent chat*): capability `agents` carries the chat's `{ min, max }`
+  down on every turn (and `said`, the person's earlier words, on `session/new`); the tree, progress,
+  approvals, findings and transcripts come up. Areas `team` (cases changed where a member used to add or
+  stop, and new ones) and `delegation`.
+
+## 3.3.0
+
+Written ahead of both implementations as LLeMbas CLI 02c7282; both carry it in the same round (`subagents.md`).
+
+- **Subagent** is the word everywhere a person or a model reads one — a `task` subagent and a
+  team's member alike; *helper* is retired. Tool names and aliases are unchanged. The `team`
+  conformance area's step `helper_done` is now **`task_done`**.
+- **The model**: a subagent runs on the session's model, unless the person named another one in
+  this session's own messages (its served id, its id at its connection or its display name, a whole
+  word, without case — never the model's text or a tool's output). `task`'s and `team_add`'s
+  `model` is refused otherwise: "Use another model for subagents only when the person names it."
+  No setting chooses a subagent's model; an agent definition's own `model` is not used for one.
+- **Capacity, the logical restrictions only**: another model on the session's model's one-model
+  server is refused ("That model is on the same one-model server as this chat's model; name a
+  model on another server."); the session's model as its own subagent needs `parallel ≥ 2` or no
+  limit, else "This model serves one request at a time: name another model, on another server,
+  for subagents." — and nothing else is picked. A member is refused for these too, and still only
+  waits for a full model.
+- **The offer**: the `agents {min, max}` setting is shown as **Subagents** (**Min**, **Max**;
+  `loop.json` `subagents.setting`) — one range from 1 to the cap with two handles, and a hint that
+  the numbers count the session's own agent: 1 is the agent alone, Max 4 the agent and up to three
+  subagents — and with it the team tools, when subagents are enabled and at
+  least one model could serve: the session's, with the capacity, or any usable model off its
+  one-model server. `context.note.team_own` tells the model, and `subagents.setting.own` the
+  person, when the session's model cannot serve subagents itself.
+- `task`'s and `team_add`'s `model` descriptions, `task`'s description (a line on the model),
+  `models/schema.json` capacity texts, `team.md` §2, §4, §8. Conformance area **`subagents`**.
+
 ## 3.2.0
 
 Written ahead of both implementations; both carry it in the same round.

harness/VERSION

@@ -1 +1 @@
-3.2.0
+3.4.0

harness/acp.md

@@ -65,6 +65,9 @@ change to the protocol is a change here first, then in both projects.
 | `rewind` | `_lembas/session/rewind` |
 | `show` | spec 3.0.0: `_lembas/show`; the `show` tool is offered in a session the client opened when the client said `show: true` in `clientCapabilities._meta.lembas` |
 | `jobs` | background work (spec 2.0.0): the `job` and `job_note` events; `_lembas/jobs`, `_lembas/job/output`, `_lembas/job/stop`, `_lembas/job/send`, `_lembas/job/transcript`; `jobs` in `_lembas/session/status`; `job` on `session/request_permission`; `machine` on a `prompt` event |
+| `progress` | spec 3.4.0 (`progress.md`): the `phase` event; `progress` in `_lembas/session/status` |
+| `agents` | spec 3.4.0: the session's Subagents setting `{ min, max }` is read from `_meta.lembas.agents` on `session/new`, `session/prompt` and `_lembas/configure` (*Subagents in an agent chat*, below) |
+| `tree` | spec 3.4.0 (`subagents.md` §5–§13): the `team` event; `_lembas/team`, `_lembas/team/transcript`, `_lembas/team/findings`, `_lembas/team/stop`; `subagent` on `session/request_permission` |
 
 ## Standard ACP, as used here
 
@@ -245,11 +248,64 @@ with options `once` (allow_once), `session` (allow_always, absent where "always"
 and `deny` (reject_once), and `_meta.lembas: { reason, purpose?, always?, patient?, job? }`. `patient`: a
 person may also answer at the terminal (a shared session), so the client does not time it out.
 `job` (capability `jobs`): `{ id, agent, description }` when a subagent running as a background job asks, so the card
-can say which — a background subagent may ask while the session itself is idle.
+can say which — a background subagent may ask while the session itself is idle. `subagent` (capability `tree`, spec 3.4.0):
+`{ id, name, position }` when the call is a tree subagent's that nobody in the tree could allow (`subagents.md` §7) —
+the card names it; the answers are the same as for the main agent's own call.
 Answer: ACP's `{ outcome: { outcome: "selected", optionId } | { outcome: "cancelled" } }`, with
 optional `_meta.lembas.command` (the corrected command line, `once` only) or
 `_meta.lembas.feedback` (why it was denied). No answer within `remote.approval_timeout` is a denial.
 
+## Subagents in an agent chat (spec 3.4.0)
+
+**An agent chat's main agent is the CLI on the device.** Its subagents are the device's: it starts
+them locally, on the models of the server it is logged in to, under every rule of `subagents.md` —
+the model rules (§2–§3), Min/Max, slots, permissions only ever narrowed, request options, and a
+call nobody in the tree can allow going to the person. The server runs no subagent of its own for
+an agent chat and offers its own subagent tool to no agent chat; what it does is carry the setting
+down and the tree up. A device without the capabilities `agents` and `tree` gets neither and runs no
+subagents for the chat (the chat says so beside the setting).
+
+**What goes down, with every turn.**
+
+- `agents: { min, max }` in `_meta.lembas` of `session/new`, of **every** `session/prompt`, and of
+  `_lembas/configure` when the person changes it between turns — the chat's setting as the server
+  keeps it (`team.md` §2: 1 ≤ min ≤ max ≤ the instance's cap). The device takes the value as the
+  session's setting for that turn: a change reaches the main agent as `team.md` §10's note, and
+  nothing else reads the device's own configuration for the chat. Absent (an older server, or a
+  client that is not LLeMbas): the device's own setting, as before 3.4.0. A value the device's own global cap is under is read as `team.md` §2
+  says (lowered, never refused).
+- `subagents: false` in the same `_meta.lembas` when the instance has subagents switched off: the
+  device offers none whatever `agents` says.
+- **What the person named** (`subagents.md` §2): the person's words reach the device as the turns'
+  prompt texts, so the device judges a named model from them as in its own sessions — every
+  `session/prompt` without `machine`; never a `machine` turn, a `_lembas/steer` from the harness, or
+  a tool's output. A chat continued on a device that has not seen its earlier turns gets them in
+  `_meta.lembas.said: string[]`, the person's earlier messages of the chat, oldest first (on
+  `session/new` only); the device counts them as typed in the session.
+- The models: the device's subagents may run on any model of the person's (`/v1/models` as the
+  device's token sees it), as in its own sessions.
+
+**What comes up.**
+
+- The tree, as it changes: the `team` event (`{ tree }`, `subagents.md` §12's data, at most four a
+  second), and on request `_lembas/team` (the tree and the whole room). Every agent's `progress`
+  is in it; the main agent's own also as the `phase` event (capability `progress`).
+- Approvals: a subagent's call that only the person can allow is an ordinary
+  `session/request_permission` with `_meta.lembas.subagent: { id, name, position }`; the server
+  shows it as its usual approval card naming the subagent, in every client. Requests the main agent
+  decides (`team_request` / `team_decide`) never reach the server as approvals: they are room
+  messages.
+- Findings and transcripts on request: `_lembas/team/findings`, `_lembas/team/transcript` (live and
+  ended subagents, read-only — the link has no method that writes into a subagent's session).
+- The person stopping a subagent: `_lembas/team/stop`.
+
+**What the server keeps.** The latest tree with the chat (so a client opening the chat draws it at
+once, and after the device has gone), and the room as `_lembas/team` last answered. Findings and
+transcripts stay on the device; a client asking for one while the device is not linked is told it
+is on the device, and gets it once the device links again. Over `/api/v2` the chat's stream carries
+the tree as it changes and the endpoints to read a subagent's transcript and findings, the same for
+a chat's own subagents and for an agent chat's — so Web, Desktop and Android draw both alike.
+
 ## `_lembas/*` methods
 
 Direction is who sends it. "→ result" marks a request; the others are notifications.
@@ -275,7 +331,7 @@ client hears the new limits.
 `{ sessionId, event }`: every engine event of the session, to a client that said `_meta.lembas`.
 `event.type` is one of `session`, `prompt`, `step`, `title`, `text`, `reasoning`,
 `tool_call_delta`, `progress`, `tool_start`, `tool_end`, `tool_denied`, `usage`, `inbox`,
-`steered`, `task`, `notice`, `retract`, `plan`, `compacted`, `sub_tool`, `job`, `job_note`, `mode`, `model`, `flash`,
+`steered`, `task`, `notice`, `retract`, `plan`, `compacted`, `sub_tool`, `job`, `job_note`, `phase`, `team`, `mode`, `model`, `flash`,
 `setting`, `error`, `done` (`src/bus/index.ts`). Those this protocol relies on:
 
 | Event | Fields |
@@ -291,6 +347,8 @@ client hears the new limits.
 | `job` | `state: "start" \| "update" \| "end"`, `job: JobInfo` (capability `jobs`) — a background job of the session began, has news (new output, a subagent's step; at most four a second per job), or ended. `JobInfo = { id, kind: "command" \| "subagent", state: "running" \| "done" \| "stopped" \| "error" \| "lost", description, command?, agent?, model?, started, ended?, exit?, unread, steps?, tools?, ownerJobId? }` — `description` what is shown for it (a command's `description` or its first line; a subagent's description), `unread` characters of output the model has not read, `exit` a command's exit code (null: a signal), `ownerJobId` set on a job a subagent started, `lost` a job that was running when the device's process ended |
 | `job_note` | `jobId`, `text` (capability `jobs`) — the note about that job was given to the model: inside a turn at the next step, or as the `machine` turn whose `prompt` follows |
 | `sub_tool` | `callId` (the `task` call the subagent belongs to), `jobId?` (its job, when it runs as one), `name`, `title`, `isError?` — a tool call of a subagent finished; a client draws it under the task line, or in the job's panel |
+| `phase` | `progress` — the progress object (`progress.md` §1), or `null` when nothing runs; `agent?` — the subagent's id when it is a subagent's, absent for the main agent's (capability `progress`). Sent when the phase changes, and for a change of its data at most four times a second. The existing `progress` event (the prompt's tokens read) is unchanged |
+| `team` | `tree` — the session's subagent tree (`subagents.md` §12, the data) — whenever it changes, at most four times a second (capability `tree`). A client draws it and fetches nothing else to do so |
 
 ### `_lembas/directories` — client → agent → result
 
@@ -301,7 +359,7 @@ Only on the link (an error on `serve --stdio`).
 
 ### `_lembas/configure` — client → agent → result
 
-Params `{ sessionId, model?: <instance's name>, effort?: <level> }` between prompts (a level as in `session/new`). Result
+Params `{ sessionId, model?: <instance's name>, effort?: <level>, agents?: { min, max } }` between prompts (a level as in `session/new`; `agents` with capability `agents`, spec 3.4.0). Result
 `{ model: <ref>, effort }`. Emits a `model` event when something changed.
 
 ### `_lembas/steer` — client → agent → result
@@ -319,9 +377,9 @@ with the message in the store; `attachments`, and the text without them) or `{ r
 ### `_lembas/session/status` — client → agent → result
 
 Params `{ sessionId }`. Result
-`{ sessionId, exists, held: "here" | "terminal" | "none", busy, title, turnId?, queued: string[], lastTurnId?, plan?, jobs? }`
+`{ sessionId, exists, held: "here" | "terminal" | "none", busy, title, turnId?, queued: string[], lastTurnId?, plan?, jobs?, progress?, tree? }`
 — `plan` the session's plan or `null` (capability `plan_state`), so a client reattaching draws it
-without waiting for the next change; `jobs` the session's background jobs, `JobInfo[]` running first (capability `jobs`), for the same reason; the turn fields with capability `turns`: the turn running now, the ones waiting, the last
+without waiting for the next change; `progress` the main agent's progress object or `null` (capability `progress`), and `tree` the subagent tree or `null` when the session has none (capability `tree`), for the same reason; `jobs` the session's background jobs, `JobInfo[]` running first (capability `jobs`), for the same reason; the turn fields with capability `turns`: the turn running now, the ones waiting, the last
 one finished (for a session not open, `held: "none"`, the last one it finished before). **Reattaching** after a restart or a dropped link, for an incomplete row with turn T:
 T running → follow it; T queued → wait, and follow it when it starts; T done → import it from
 `history`; T unknown → send the prompt again (safe: the device deduplicates).
@@ -362,6 +420,30 @@ with its output — so a client can show the subagent as a panel and follow it w
 `job` events that come after. `-32602` for a command job. A subagent's transcript lasts as long as
 its session row on the device: it is kept with the parent session and deleted with it.
 
+### `_lembas/team` — client → agent → result (capability `tree`)
+
+Params `{ sessionId, since?: number }`. Result `{ tree, room }`: the subagent tree (`subagents.md`
+§12) and the room's messages numbered above `since` (default 0), every one — the person sees the
+whole room — each `{ n, from, to, recipients, text, time, report, kind }` with `kind` one of
+`message`, `report`, `assignment`, `request`, `decision`. `tree` is `null` for a session with no
+subagents. For a session held by a terminal, answered through the hub, as `status` is.
+
+### `_lembas/team/transcript` — client → agent → result (capability `tree`)
+
+Params `{ sessionId, id }` — a subagent's id (`s1`), live or ended. Result `{ id, state, turns }` in
+the shape of `_lembas/session/history`'s `turns`: its conversation so far, **read-only** — the link
+has no method to write into it. A client following it live refetches on the `team` event and draws
+the `sub_tool` events meanwhile. `-32602` for an id the session never had.
+
+### `_lembas/team/findings` — client → agent → result (capability `tree`)
+
+Params `{ sessionId, id }`. Result `{ id, findings: [{ n, text, time }] }`, ended subagents' too.
+
+### `_lembas/team/stop` — client → agent → result (capability `tree`)
+
+Params `{ sessionId, id }`: the person stops a subagent and its subtree (`subagents.md` §10), said to
+the main agent as a note. Result `{ stopped: [id, …] }`; `-32602` for an id that is not live.
+
 ### `_lembas/session/delete` — client → agent → result
 
 Params `{ sessionId }`: the chat was deleted. The session stops if it works, leaves the store (its

harness/conformance/delegation.json

@@ -0,0 +1,433 @@
+{
+  "about": "Delegation (subagents.md §7, spec 3.4.0): a subagent's mode and request options, and who decides its calls. The project root is /project, nothing named exists on disk, and only the default rules apply (permission/defaults.json). `add` — team_add's mode check for a parent in `parent_mode` with `mode` and `request` as given → `{result, mode}`: the refusal word for word, or null and the mode the subagent gets. `call` — a subagent in `mode` with `request` options and `grants` (rules {tool, pattern} the main agent allowed) under a main agent in `main_mode`, calling `call` (`bash` with a `line`, or a file tool with a `path` relative to the root) → `{answer, text?}`: `run`, `person` (an approval card to the person), or `refused` with its text, word for word — except where the case says `own_reason` (a hardline's, a mode's or the unattended refusal's reason is each project's), when only the answer is compared. `unattended`: nobody can answer. `request` — team_request for `tool` by a subagent with `request` options, its first → `{result}`. `switch` — the main agent's mode changes from `main_from` to `main_to` with these subagents (each with its parent and mode) → each one's mode after. LLeMbas: services/team.py; CLI: team/delegation.ts.",
+  "cases": [
+    {
+      "about": "No mode given: the parent's.",
+      "op": "add",
+      "parent_mode": "edit",
+      "expect": {
+        "result": null,
+        "mode": "edit"
+      }
+    },
+    {
+      "about": "Stricter is allowed.",
+      "op": "add",
+      "parent_mode": "edit",
+      "mode": "manual",
+      "expect": {
+        "result": null,
+        "mode": "manual"
+      }
+    },
+    {
+      "about": "Looser is refused, naming the parent's mode.",
+      "op": "add",
+      "parent_mode": "manual",
+      "mode": "edit",
+      "expect": {
+        "result": "A subagent's mode cannot be looser than its parent's (manual).",
+        "mode": null
+      }
+    },
+    {
+      "about": "Plan is the strictest: anything else under it is looser.",
+      "op": "add",
+      "parent_mode": "plan",
+      "mode": "auto",
+      "expect": {
+        "result": "A subagent's mode cannot be looser than its parent's (plan).",
+        "mode": null
+      }
+    },
+    {
+      "about": "Request options with the parent's own mode: nothing to ask for.",
+      "op": "add",
+      "parent_mode": "auto",
+      "mode": "auto",
+      "request": [
+        "write"
+      ],
+      "expect": {
+        "result": "Request options go with a mode stricter than the parent's: without one there is nothing to ask for.",
+        "mode": null
+      }
+    },
+    {
+      "about": "Request options with no mode given (the parent's): refused the same way.",
+      "op": "add",
+      "parent_mode": "edit",
+      "request": [
+        "execute"
+      ],
+      "expect": {
+        "result": "Request options go with a mode stricter than the parent's: without one there is nothing to ask for.",
+        "mode": null
+      }
+    },
+    {
+      "about": "Request options with a stricter mode.",
+      "op": "add",
+      "parent_mode": "edit",
+      "mode": "manual",
+      "request": [
+        "write",
+        "execute"
+      ],
+      "expect": {
+        "result": null,
+        "mode": "manual"
+      }
+    },
+    {
+      "about": "An unknown kind is refused, naming it.",
+      "op": "add",
+      "parent_mode": "edit",
+      "mode": "manual",
+      "request": [
+        "delete"
+      ],
+      "expect": {
+        "result": "delete is not a request option: use read, write or execute.",
+        "mode": null
+      }
+    },
+    {
+      "about": "An empty list of request options is no options, and needs nothing.",
+      "op": "add",
+      "parent_mode": "edit",
+      "request": [],
+      "expect": {
+        "result": null,
+        "mode": "edit"
+      }
+    },
+    {
+      "about": "All in edit: an edit inside the project runs.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "edit",
+      "call": {
+        "tool": "edit",
+        "path": "src/a.ts"
+      },
+      "expect": {
+        "answer": "run"
+      }
+    },
+    {
+      "about": "All in edit: bash that edit mode asks goes to the person — the main agent could not allow it either.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "edit",
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "expect": {
+        "answer": "person"
+      }
+    },
+    {
+      "about": "All in manual: an edit goes to the person.",
+      "op": "call",
+      "main_mode": "manual",
+      "mode": "manual",
+      "call": {
+        "tool": "edit",
+        "path": "src/a.ts"
+      },
+      "expect": {
+        "answer": "person"
+      }
+    },
+    {
+      "about": "Narrowed with no request options: what the main agent could allow is refused.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "manual",
+      "call": {
+        "tool": "edit",
+        "path": "src/a.ts"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, and you may not ask for it: tell your parent what you need."
+      }
+    },
+    {
+      "about": "Narrowed with the kind in its options: it asks the main agent.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "manual",
+      "request": [
+        "write"
+      ],
+      "call": {
+        "tool": "edit",
+        "path": "src/a.ts"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, but the main agent may: ask for it with team_request (the tool, the pattern, why), then try again once it allows."
+      }
+    },
+    {
+      "about": "Narrowed, the kind not in its options.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "edit",
+      "request": [
+        "write"
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, and you may not ask for it: tell your parent what you need."
+      }
+    },
+    {
+      "about": "Narrowed, execute in its options: asks the main agent.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "edit",
+      "request": [
+        "execute"
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, but the main agent may: ask for it with team_request (the tool, the pattern, why), then try again once it allows."
+      }
+    },
+    {
+      "about": "A grant from the main agent lets the call run.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "edit",
+      "request": [
+        "execute"
+      ],
+      "grants": [
+        {
+          "tool": "bash",
+          "pattern": "git push *"
+        }
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "expect": {
+        "answer": "run"
+      }
+    },
+    {
+      "about": "A grant for another pattern does not.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "edit",
+      "request": [
+        "execute"
+      ],
+      "grants": [
+        {
+          "tool": "bash",
+          "pattern": "git push *"
+        }
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "git commit -m x"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, but the main agent may: ask for it with team_request (the tool, the pattern, why), then try again once it allows."
+      }
+    },
+    {
+      "about": "A grant never reaches past the main agent's own mode: what it would have to ask goes to the person.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "manual",
+      "request": [
+        "execute"
+      ],
+      "grants": [
+        {
+          "tool": "bash",
+          "pattern": "git push *"
+        }
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "expect": {
+        "answer": "person"
+      }
+    },
+    {
+      "about": "Reading .env asks in every mode: it goes to the person.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "edit",
+      "call": {
+        "tool": "read",
+        "path": ".env"
+      },
+      "expect": {
+        "answer": "person"
+      }
+    },
+    {
+      "about": "Reading .env under an auto main agent: a narrowed subagent with read in its options asks it.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "manual",
+      "request": [
+        "read"
+      ],
+      "call": {
+        "tool": "read",
+        "path": ".env"
+      },
+      "expect": {
+        "answer": "refused",
+        "text": "Your mode does not allow this, but the main agent may: ask for it with team_request (the tool, the pattern, why), then try again once it allows."
+      }
+    },
+    {
+      "about": "Plan mode's refusal is final: request options do not reach past it.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "plan",
+      "request": [
+        "execute"
+      ],
+      "call": {
+        "tool": "bash",
+        "line": "rm -rf build"
+      },
+      "own_reason": true,
+      "expect": {
+        "answer": "refused"
+      }
+    },
+    {
+      "about": "What the rules allow runs in every mode.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "plan",
+      "call": {
+        "tool": "bash",
+        "line": "git status"
+      },
+      "expect": {
+        "answer": "run"
+      }
+    },
+    {
+      "about": "The hardline refuses in every mode, auto included.",
+      "op": "call",
+      "main_mode": "auto",
+      "mode": "auto",
+      "call": {
+        "tool": "bash",
+        "line": "rm -rf /"
+      },
+      "own_reason": true,
+      "expect": {
+        "answer": "refused"
+      }
+    },
+    {
+      "about": "Unattended: what would go to the person is refused.",
+      "op": "call",
+      "main_mode": "edit",
+      "mode": "edit",
+      "unattended": true,
+      "call": {
+        "tool": "bash",
+        "line": "git push origin main"
+      },
+      "own_reason": true,
+      "expect": {
+        "answer": "refused"
+      }
+    },
+    {
+      "about": "team_request: a tool whose risk class is not in the options is refused, naming them.",
+      "op": "request",
+      "request": [
+        "write"
+      ],
+      "tool": "bash",
+      "expect": {
+        "result": "You may not ask for bash: your request options are write."
+      }
+    },
+    {
+      "about": "team_request: a kind in the options is asked, with the request's id.",
+      "op": "request",
+      "request": [
+        "write",
+        "execute"
+      ],
+      "tool": "bash",
+      "expect": {
+        "result": "Asked the main agent (r1): you are told its decision. Carry on with what you can meanwhile."
+      }
+    },
+    {
+      "about": "The main agent made stricter: every subagent looser than its parent is made as strict as it, down the tree; a stricter one keeps its own.",
+      "op": "switch",
+      "agents": [
+        {
+          "name": "api",
+          "parent": "lead",
+          "mode": "auto"
+        },
+        {
+          "name": "tests",
+          "parent": "api",
+          "mode": "edit"
+        },
+        {
+          "name": "docs",
+          "parent": "lead",
+          "mode": "plan"
+        }
+      ],
+      "main_from": "auto",
+      "main_to": "manual",
+      "expect": {
+        "api": "manual",
+        "tests": "manual",
+        "docs": "plan"
+      }
+    },
+    {
+      "about": "The main agent made looser: no subagent changes.",
+      "op": "switch",
+      "agents": [
+        {
+          "name": "api",
+          "parent": "lead",
+          "mode": "edit"
+        }
+      ],
+      "main_from": "edit",
+      "main_to": "auto",
+      "expect": {
+        "api": "edit"
+      }
+    }
+  ]
+}

harness/conformance/progress.json

@@ -0,0 +1,333 @@
+{
+  "about": "Progress (progress.md, spec 3.4.0): the text and the line every client draws for a progress object. `progress`: the object; `now`: the client's clock (Unix ms). Expect `text` — the phase's text from loop.json progress.texts, its fields filled (numbers written plainly: integers as they are, a rate with one decimal) — and `line` — `{{text}} · {{elapsed}}` with elapsed = now − since in whole seconds (5s, 1m 05s, 1h 02m), or the text alone for stopped and failed. retrying's `in` is counted down: in − floor((now − since) / 1000), never below 0. LLeMbas: services/progress.py; CLI: session/progress.ts.",
+  "cases": [
+    {
+      "about": "Queued, nothing more known.",
+      "progress": {
+        "phase": "queued",
+        "since": 1762790400000
+      },
+      "now": 1762790405000,
+      "expect": {
+        "text": "Queued",
+        "line": "Queued · 5s"
+      }
+    },
+    {
+      "about": "Queued behind two turns.",
+      "progress": {
+        "phase": "queued",
+        "since": 1762790400000,
+        "ahead": 2
+      },
+      "now": 1762790465000,
+      "expect": {
+        "text": "Queued: 2 ahead",
+        "line": "Queued: 2 ahead · 1m 05s"
+      }
+    },
+    {
+      "about": "A one-request server answering another request.",
+      "progress": {
+        "phase": "waiting_for_model",
+        "since": 1762790400000,
+        "model": "Qwen 27B",
+        "reason": "busy"
+      },
+      "now": 1762790412400,
+      "expect": {
+        "text": "Waiting for Qwen 27B: it serves one request at a time and is answering another",
+        "line": "Waiting for Qwen 27B: it serves one request at a time and is answering another · 12s"
+      }
+    },
+    {
+      "about": "Every slot taken.",
+      "progress": {
+        "phase": "waiting_for_model",
+        "since": 1762790400000,
+        "model": "coder",
+        "reason": "slots"
+      },
+      "now": 1762790401000,
+      "expect": {
+        "text": "Waiting for coder: all its slots are in use",
+        "line": "Waiting for coder: all its slots are in use · 1s"
+      }
+    },
+    {
+      "about": "Another model on the one-model server.",
+      "progress": {
+        "phase": "waiting_for_model",
+        "since": 1762790400000,
+        "model": "coder",
+        "reason": "group"
+      },
+      "now": 1762790400000,
+      "expect": {
+        "text": "Waiting for coder: its server holds another model now",
+        "line": "Waiting for coder: its server holds another model now · 0s"
+      }
+    },
+    {
+      "about": "Loading.",
+      "progress": {
+        "phase": "loading_model",
+        "since": 1762790400000,
+        "model": "Gemma 12B"
+      },
+      "now": 1762794125000,
+      "expect": {
+        "text": "Loading Gemma 12B",
+        "line": "Loading Gemma 12B · 1h 02m"
+      }
+    },
+    {
+      "about": "Reading the prompt: the percentage is floored.",
+      "progress": {
+        "phase": "reading_prompt",
+        "since": 1762790400000,
+        "done": 1999,
+        "total": 3000
+      },
+      "now": 1762790402000,
+      "expect": {
+        "text": "Reading the prompt: 1999 / 3000 tokens (66%)",
+        "line": "Reading the prompt: 1999 / 3000 tokens (66%) · 2s"
+      }
+    },
+    {
+      "about": "Reading the prompt with no counts from the server.",
+      "progress": {
+        "phase": "reading_prompt",
+        "since": 1762790400000
+      },
+      "now": 1762790402000,
+      "expect": {
+        "text": "Reading the prompt",
+        "line": "Reading the prompt · 2s"
+      }
+    },
+    {
+      "about": "Thinking.",
+      "progress": {
+        "phase": "thinking",
+        "since": 1762790400000
+      },
+      "now": 1762790459999,
+      "expect": {
+        "text": "Thinking",
+        "line": "Thinking · 59s"
+      }
+    },
+    {
+      "about": "Writing, the rate with one decimal.",
+      "progress": {
+        "phase": "writing",
+        "since": 1762790400000,
+        "rate": 12.46
+      },
+      "now": 1762790404000,
+      "expect": {
+        "text": "Writing: 12.5 tok/s",
+        "line": "Writing: 12.5 tok/s · 4s"
+      }
+    },
+    {
+      "about": "Writing before a rate is measured.",
+      "progress": {
+        "phase": "writing",
+        "since": 1762790400000
+      },
+      "now": 1762790400400,
+      "expect": {
+        "text": "Writing",
+        "line": "Writing · 0s"
+      }
+    },
+    {
+      "about": "A tool with what it does.",
+      "progress": {
+        "phase": "running_tool",
+        "since": 1762790400000,
+        "tool": "bash",
+        "what": "bun test"
+      },
+      "now": 1762790461000,
+      "expect": {
+        "text": "Running bash: bun test",
+        "line": "Running bash: bun test · 1m 01s"
+      }
+    },
+    {
+      "about": "A tool without a title.",
+      "progress": {
+        "phase": "running_tool",
+        "since": 1762790400000,
+        "tool": "recall"
+      },
+      "now": 1762790401000,
+      "expect": {
+        "text": "Running recall",
+        "line": "Running recall · 1s"
+      }
+    },
+    {
+      "about": "An approval, naming the tool.",
+      "progress": {
+        "phase": "waiting_for_person",
+        "since": 1762790400000,
+        "what": "approval",
+        "tool": "bash"
+      },
+      "now": 1762790430000,
+      "expect": {
+        "text": "Waiting for you: allow bash?",
+        "line": "Waiting for you: allow bash? · 30s"
+      }
+    },
+    {
+      "about": "An approval with no tool named.",
+      "progress": {
+        "phase": "waiting_for_person",
+        "since": 1762790400000,
+        "what": "approval"
+      },
+      "now": 1762790400000,
+      "expect": {
+        "text": "Waiting for you: an approval",
+        "line": "Waiting for you: an approval · 0s"
+      }
+    },
+    {
+      "about": "A question.",
+      "progress": {
+        "phase": "waiting_for_person",
+        "since": 1762790400000,
+        "what": "question"
+      },
+      "now": 1762790403000,
+      "expect": {
+        "text": "Waiting for you: a question",
+        "line": "Waiting for you: a question · 3s"
+      }
+    },
+    {
+      "about": "The plan.",
+      "progress": {
+        "phase": "waiting_for_person",
+        "since": 1762790400000,
+        "what": "plan"
+      },
+      "now": 1762790403000,
+      "expect": {
+        "text": "Waiting for you: the plan",
+        "line": "Waiting for you: the plan · 3s"
+      }
+    },
+    {
+      "about": "Subagents.",
+      "progress": {
+        "phase": "waiting_for_subagents",
+        "since": 1762790400000,
+        "running": 3
+      },
+      "now": 1762790490000,
+      "expect": {
+        "text": "Waiting for subagents: 3 running",
+        "line": "Waiting for subagents: 3 running · 1m 30s"
+      }
+    },
+    {
+      "about": "Compacting.",
+      "progress": {
+        "phase": "compacting",
+        "since": 1762790400000
+      },
+      "now": 1762790407000,
+      "expect": {
+        "text": "Compacting the conversation",
+        "line": "Compacting the conversation · 7s"
+      }
+    },
+    {
+      "about": "Retrying: `in` counts down from since.",
+      "progress": {
+        "phase": "retrying",
+        "since": 1762790400000,
+        "reason": "502 Bad Gateway",
+        "in": 10,
+        "attempt": 2
+      },
+      "now": 1762790403000,
+      "expect": {
+        "text": "Retrying in 7 s: 502 Bad Gateway",
+        "line": "Retrying in 7 s: 502 Bad Gateway · 3s"
+      }
+    },
+    {
+      "about": "Retrying: never below 0.",
+      "progress": {
+        "phase": "retrying",
+        "since": 1762790400000,
+        "reason": "overloaded",
+        "in": 2
+      },
+      "now": 1762790405000,
+      "expect": {
+        "text": "Retrying in 0 s: overloaded",
+        "line": "Retrying in 0 s: overloaded · 5s"
+      }
+    },
+    {
+      "about": "Stopped draws no elapsed time.",
+      "progress": {
+        "phase": "stopped",
+        "since": 1762790400000
+      },
+      "now": 1762790405000,
+      "expect": {
+        "text": "Stopped",
+        "line": "Stopped"
+      }
+    },
+    {
+      "about": "Stopped by the person.",
+      "progress": {
+        "phase": "stopped",
+        "since": 1762790400000,
+        "reason": "person"
+      },
+      "now": 1762790405000,
+      "expect": {
+        "text": "Stopped: by you",
+        "line": "Stopped: by you"
+      }
+    },
+    {
+      "about": "Stopped at the step limit.",
+      "progress": {
+        "phase": "stopped",
+        "since": 1762790400000,
+        "reason": "steps"
+      },
+      "now": 1762790400000,
+      "expect": {
+        "text": "Stopped: step limit",
+        "line": "Stopped: step limit"
+      }
+    },
+    {
+      "about": "Failed.",
+      "progress": {
+        "phase": "failed",
+        "since": 1762790400000,
+        "reason": "connection refused"
+      },
+      "now": 1762790405000,
+      "expect": {
+        "text": "Failed: connection refused",
+        "line": "Failed: connection refused"
+      }
+    }
+  ]
+}

harness/loop.json

@@ -34,7 +34,7 @@
     "about": "Calls in one step run together, at most 4; a call that waits on the user (ask_user, plan_submit) runs alone; two calls on one file run in order. A tool's error is its result, never the end of the prompt.",
     "max": 4
   },
-  "capacity": "A subagent (task) is not started on a model whose server cannot hold it beside the prompt's own model — or beside the model of a subagent already running: another model of a connection that holds one model at a time, or the same model when it already serves as many of this session's requests as it can at once (models/schema.json capacity: `parallel`, `single_session` being parallel 1). A team's members are not refused there: they wait for a slot (team.md, Capacity).",
+  "capacity": "A subagent (task) is not started on a model whose server cannot hold it beside the prompt's own model — or beside the model of a subagent already running: another model of a connection that holds one model at a time, or the same model when it already serves as many of this session's requests as it can at once (models/schema.json capacity: `parallel`, `single_session` being parallel 1). Spec 3.3.0: which model a subagent may run on at all, and the refusals for the session's own one-model server and a one-request model, are subagents.md and `subagents` below, checked first. A team's members are not refused for a full model: they wait for a slot (team.md, Capacity).",
   "jobs": {
     "about": "Background work owned by a session: a command (bash `background: true`, or one still running when its timeout passed) or a subagent (task `background: true`). One id sequence per session (job1, job2, …), one list (job_list), one way to read (job_output), to stop (job_stop) and, for a subagent, to message (job_send); a subagent's own tools see only the jobs it started. A job that ends wakes the session that owns it: while a prompt runs, as a user message at the next step boundary; otherwise as a new turn of its own, marked `machine` so a UI shows it as the harness's, not the person's. A session that is not open keeps the note until it is resumed. Every such message begins with its mark, said rather than hidden: a model refuses an unmarked one as an injection (conformance/jobs.json gives the texts).",
     "marks": {
@@ -61,8 +61,30 @@
     },
     "lifecycle": "A subagent's own background commands stop when it ends, and its report says so. A session's jobs stop when the session is deleted. A front end asked to exit while jobs run asks the person first; one with nobody to ask waits for the jobs the prompt started, then stops the rest."
   },
+  "subagents": {
+    "about": "Spec 3.3.0 (subagents.md): which model a subagent — a task subagent or a team member — runs on, and when subagents are offered. The session's own model unless the person named another in this session; no setting chooses one. The texts are said word for word, checked in this order: not_named, same_group, one_request. conformance/subagents.json.",
+    "texts": {
+      "not_named": "Use another model for subagents only when the person names it.",
+      "same_group": "That model is on the same one-model server as this chat's model; name a model on another server.",
+      "one_request": "This model serves one request at a time: name another model, on another server, for subagents."
+    },
+    "named": {
+      "about": "A model is named when its served id, its id at its connection or its display name occurs as a whole word, without case, in a message the person typed in this session (never model text, tool output, files, notes, marked messages or reports; a compacted message still counts). Before the name: not a letter, digit, _, -, /, . or :. After it: not a letter, digit, _, - or /, nor a . or : followed by a letter or digit. White space in a display name matches any run of white space.",
+      "before": "[\\p{L}\\p{N}_\\-/.:]",
+      "after": "[\\p{L}\\p{N}_\\-/]|[.:][\\p{L}\\p{N}]"
+    },
+    "setting": {
+      "about": "How the agents {min, max} setting (team.md §2) is shown to a person: one range from 1 to the cap with two handles, `min` and `max` naming them; `label` beside it, `hint` (or `short` where one line has no room) saying what the numbers count — the session's own agent included, so 1 is that agent alone and Max 4 is it and up to 3 subagents; and `own` beside it when the session's model cannot serve subagents itself (subagents.md §4).",
+      "label": "Subagents",
+      "min": "Min",
+      "max": "Max",
+      "hint": "The count includes this agent: 1 is this agent alone, with no subagents; Max 4 is this agent and up to 3 subagents.",
+      "short": "1 is this agent alone",
+      "own": "This model serves one request at a time: subagents need another model, on another server, named in the chat."
+    }
+  },
   "team": {
-    "about": "Teams (spec 3.2.0, team.md): a lead and the members it adds, who talk in the team's room. The texts are said word for word; {{size}} is the team's size after the call (team.md §3), {{names}} names joined with \", \" in tree order. conformance/team.json.",
+    "about": "Teams (spec 3.2.0, team.md; which model a member runs on and when a team is offered: spec 3.3.0, subagents.md; the main agent, the tree, permissions and a subagent's life: spec 3.4.0, subagents.md §5–§13). The texts are said word for word; {{size}} is the team's size after the call (team.md §3), {{names}} names joined with \", \" in tree order. conformance/team.json, conformance/delegation.json.",
     "agents": {
       "about": "The chat setting agents {min, max}: the agents a team may have, the lead included. 1 ≤ min ≤ max ≤ cap. min and max are a new chat's defaults; cap is the instance's (LLeMbas) or the global configuration's (the CLI).",
       "min": 1,
@@ -70,20 +92,27 @@
       "cap": 32
     },
     "names": {
-      "about": "A member's name: unique in the team; `lead` is the lead's, `all` addresses everyone.",
+      "about": "A subagent's name: unique among the subagents that have not ended; `lead` is the main agent's, `all` addresses everyone, `parent` (3.4.0) the sender's parent.",
       "pattern": "^[a-z][a-z0-9-]{0,31}$",
       "lead": "lead",
       "reserved": [
         "all",
-        "lead"
+        "lead",
-      ]
+        "parent"
+      ],
+      "parent": "parent",
+      "id_prefix": "s"
     },
     "marks": {
       "message": "[Team message from {{from}} to {{to}}:]",
-      "report": "[Team report from {{from}} to {{to}}:]"
+      "report": "[Team report from {{from}} to {{to}}:]",
+      "assignment": "[Team assignment from {{from}} to {{to}}:]",
+      "request": "[Team request {{id}} from {{from}} to {{to}}:]",
+      "decision": "[Team decision on {{id}} from {{from}} to {{to}}:]",
+      "context_full": "[Team notice to {{to}}: your context is full]"
     },
     "unavailable": {
-      "about": "What a team tool answers while its requirement `team` is not met (team.md §2). one_agent is also each team tool's own `unavailable`, and what task answers while a team is offered, max is 1 and the lead has no member.",
+      "about": "What a team tool answers while its requirement `team` is not met (team.md §2). one_agent is also each team tool's own `unavailable`, and what task answers while a team is offered, max is 1 and the lead has no member. one_request: no model can serve as a subagent any more (subagents.md §4) — the session's model's capacity changed after the prompt was made.",
       "one_agent": "The person allows one agent in this chat: there is no team. Do the work yourself.",
       "one_request": "This model serves one request at a time now, so there is no team. Do the work yourself."
     },
@@ -131,25 +160,106 @@
       "not_yours": "{{name}} is not under you: you can stop only the members you lead, directly or through your members.",
       "nobody_to_stop": "No member is under you: there is nobody to stop.",
       "no_report": "(no report)",
-      "ended": "[ended: {{reason}}]"
+      "ended": "[ended: {{reason}}]",
+      "only_main": "Only the main agent adds, assigns, replaces and stops subagents: ask it (team_send to lead).",
+      "not_manager": "{{name}} is not a sub-manager: place a subagent under the main agent (lead) or under one added with lead: true.",
+      "wider": "A subagent's mode cannot be looser than its parent's ({{parent_mode}}).",
+      "request_not_narrower": "Request options go with a mode stricter than the parent's: without one there is nothing to ask for.",
+      "request_kind": "{{kind}} is not a request option: use read, write or execute.",
+      "requestable": "Your mode does not allow this, but the main agent may: ask for it with team_request (the tool, the pattern, why), then try again once it allows.",
+      "not_delegated": "Your mode does not allow this, and you may not ask for it: tell your parent what you need.",
+      "not_requestable": "You may not ask for {{tool}}: your request options are {{kinds}}.",
+      "requested": "Asked the main agent ({{id}}): you are told its decision. Carry on with what you can meanwhile.",
+      "no_request": "No open request {{id}}.",
+      "decided": "{{decision}} {{id}} for {{name}}.",
+      "allowed": "Allowed",
+      "refused": "Refused",
+      "decision_allowed": "Allowed: {{tool}} {{pattern}}. Try it again now.",
+      "decision_refused": "Refused: {{tool}} {{pattern}}.",
+      "no_parent": "You are the main agent: you have no parent.",
+      "finished": "Reported to {{to}}. Your task is done: stop here.",
+      "has_ended": "{{name}} has ended: replace it (team_replace) or add another.",
+      "cycle": "{{name}} cannot go under {{parent}}: that is under it.",
+      "assigned": "Assigned to {{name}}: it is told now.",
+      "replaced": "Replaced {{name}}: {{id}} takes its place, briefed with its task and findings. The team is {{size}} of {{max}} agents.",
+      "replace_brief": "{{task}}\n\nYou take over from {{name}} ({{id}}), which ended. What it found:\n{{findings}}",
+      "no_findings": "(no findings)",
+      "noted": "Noted (finding {{n}}).",
+      "not_yours_findings": "You may read only your own findings and those of the subagents under you.",
+      "context_full": "Save what you found with team_findings, then tell your parent with team_send that you are gone and what is left undone. This is your last turn: at most {{steps}} steps.",
+      "gone": "{{name}} ({{id}}) ran out of context and has ended. Its findings: team_findings with of {{name}}.",
+      "request_body": "{{tool}} {{pattern}} — {{why}}"
     },
     "read": {
-      "about": "team_read's answer: the tree, then the room. Agents in tree order (an agent, then its members in the order they were added, depth first), indented two spaces per level below the lead. `last`: `active now` while working, else how long ago its last turn ended (durations as jobs write them: \"5s ago\", \"1m 05s ago\"). A message's text: its first line after the colon, every further line indented two spaces. Messages above `since` the caller may see, oldest first, at most read_max; `more` when there are others after the last shown.",
+      "about": "team_read's answer: the tree, then the room. Agents in tree order (an agent, then its members in the order they were added, depth first), indented two spaces per level below the lead. `last`: `active now` while working, else how long ago its last turn ended (durations as jobs write them: \"5s ago\", \"1m 05s ago\"). A message's text: its first line after the colon, every further line indented two spaces. Messages above `since` the caller may see, oldest first, at most read_max; `more` when there are others after the last shown. Spec 3.4.0: the row adds the id, the mode (with request options) and the context use; open slots follow the tree, ended subagents after them.",
       "tree": "Team ({{size}} of {{max}} agents):",
-      "agent": "{{indent}}- {{name}} — {{role}} · {{state}} · {{model}} · {{last}}",
+      "agent": "{{indent}}- {{name}} ({{id}}) — {{role}} · {{state}} · {{model}} · {{mode}} · {{context}} · {{last}}",
-      "lead_role": "leads the team",
+      "lead_role": "the main agent",
       "states": {
         "working": "working",
         "waiting": "waiting for a slot",
         "finished": "finished",
-        "idle": "idle"
+        "idle": "idle",
+        "paused": "paused: context full",
+        "done": "done",
+        "out_of_context": "out of context",
+        "stopped": "stopped"
       },
       "active": "active now",
       "room": "Room since #{{since}}:",
       "message": "#{{n}} {{from}} → {{to}}: {{text}}",
       "report": "#{{n}} {{from}} → {{to}} (report): {{text}}",
       "none": "No messages since #{{since}}.",
-      "more": "More after #{{last}}: team_read with since {{last}}."
+      "more": "More after #{{last}}: team_read with since {{last}}.",
+      "slots": "Open slots: {{n}} — fill them with team_add.",
+      "ended": "Ended:",
+      "ended_agent": "- {{name}} ({{id}}) — {{role}} · {{state}} · {{findings}} findings",
+      "context": "{{percent}}% context",
+      "request": "+ {{kinds}}"
+    },
+    "main_only": {
+      "about": "Spec 3.4.0 (subagents.md §6, §8): the tools only the main agent has; a subagent's call to one, or to task, answers texts.only_main.",
+      "tools": [
+        "team_add",
+        "team_assign",
+        "team_decide",
+        "team_replace",
+        "team_stop"
+      ]
+    },
+    "context": {
+      "about": "Spec 3.4.0 (subagents.md §9): a subagent does not compact. When its context reaches pause_at of its model's window it is paused and given one final turn of at most final_steps steps, whose user message is the context_full mark and texts.context_full.",
+      "pause_at": 0.85,
+      "final_steps": 8
+    },
+    "request": {
+      "about": "Spec 3.4.0 (subagents.md §7): the kinds a request option may name — risk classes — and the ids of requests, r1, r2, … per session.",
+      "kinds": [
+        "read",
+        "write",
+        "execute"
+      ],
+      "id_prefix": "r"
+    },
+    "whoami": {
+      "about": "Spec 3.4.0 (subagents.md §8): team_whoami's answer, one line each; `request` only with request options, `parent` only for a subagent. {{percent}} is floor(used × 100 / window).",
+      "lines": {
+        "who": "You are {{name}} ({{id}}): {{role}}.",
+        "parent": "Your parent: {{parent}}.",
+        "task": "Your task: {{task}}",
+        "mode": "Your mode: {{mode}}.",
+        "request": "You may ask the main agent for: {{kinds}} (team_request).",
+        "model": "Your model: {{model}}.",
+        "limits": "Limits: the team is {{size}} of {{max}} agents; at most {{steps}} steps a turn; {{spawn}}",
+        "spawn_yes": "you add subagents.",
+        "spawn_no": "only the main agent adds subagents.",
+        "context": "Context: {{used}} of {{window}} tokens ({{percent}}%)."
+      }
+    },
+    "findings": {
+      "about": "Spec 3.4.0: team_findings read's answer — the header, then one line per finding, oldest first; none: texts.no_findings after the header.",
+      "header": "Findings of {{name}} ({{id}}):",
+      "item": "{{n}}. {{text}}"
     }
   },
   "context": {
@@ -203,7 +313,7 @@
       "started": "Started in: {{path}} (relative to the project root)"
     },
     "note": {
-      "about": "A note on the next user message (context.md §10, rule 3): what changed since the system prompt was made, after the person's words. Lines in this order, each only when it applies. `plan_ended` is the mode block for a session that leaves plan mode for a mode with no text of its own (a chat). `plan_own` is the plan block for a session that does not resend earlier tool results (a chat): the model's own plan, as its tools left it. `personal` is the {{path}} of the instructions block for the person's own instructions. Spec 3.2.0, where a team is offered (team.md §10): `team` — the size setting, on the first user message and after it changes, followed in the same block by `team_lead` when min ≥ 2 ({{members}} = min − 1), or `team_one` in its place when max is 1; `team_stopped` — members the person stopped, a block of its own.",
+      "about": "A note on the next user message (context.md §10, rule 3): what changed since the system prompt was made, after the person's words. Lines in this order, each only when it applies. `plan_ended` is the mode block for a session that leaves plan mode for a mode with no text of its own (a chat). `plan_own` is the plan block for a session that does not resend earlier tool results (a chat): the model's own plan, as its tools left it. `personal` is the {{path}} of the instructions block for the person's own instructions. Spec 3.2.0, where a team is offered (team.md §10): `team` — the size setting, on the first user message and after it changes, followed in the same block by `team_lead` when min ≥ 2 ({{members}} = min − 1), or `team_one` in its place when max is 1; `team_stopped` — members the person stopped, a block of its own. Spec 3.3.0 (subagents.md §4): `team_own` after `team` (and after `team_lead`) in the same block when the session's model cannot serve subagents itself — it serves one request at a time — and another server's model could.",
       "open": "<session-note>",
       "close": "</session-note>",
       "date": "Today is {{date}}.",
@@ -217,11 +327,116 @@
       "personality": "The person chose a different personality:\n{{personality}}",
       "mcp": "Instructions from an MCP server that connected since the session began:\n{{instructions}}",
       "team": "The team may have {{min}} to {{max}} agents, you included.",
-      "team_lead": "Work through a team from the start: you lead, and members do the work — at least {{members}} of them.",
+      "team_lead": "Work through subagents from the start: you lead, and they do the work. The person opened {{members}} slots for them: fill them with team_add.",
       "team_one": "The person allows one agent: work alone; the team tools and task are unavailable.",
+      "team_own": "This model serves one request at a time, so subagents cannot run on it: they need another model, on another server, that the person names in this chat.",
       "team_stopped": "The person stopped {{names}}; they have left the team.",
       "empty": "(none)"
     },
     "mcp_wait_ms": 3000
+  },
+  "progress": {
+    "about": "Spec 3.4.0 (progress.md): one fixed set of phases every client shows the same way. `texts` are the shared i18n keys progress.<phase> (a variant: progress.<phase>.<variant>); `line` is drawn for every phase but stopped and failed; elapsed in whole seconds (5s, 1m 05s, 1h 02m). conformance/progress.json.",
+    "phases": [
+      "queued",
+      "waiting_for_model",
+      "loading_model",
+      "reading_prompt",
+      "thinking",
+      "writing",
+      "running_tool",
+      "waiting_for_person",
+      "waiting_for_subagents",
+      "compacting",
+      "retrying",
+      "stopped",
+      "failed"
+    ],
+    "fields": {
+      "queued": {
+        "ahead": "int?"
+      },
+      "waiting_for_model": {
+        "model": "string",
+        "reason": "busy|slots|group"
+      },
+      "loading_model": {
+        "model": "string"
+      },
+      "reading_prompt": {
+        "done": "int?",
+        "total": "int?"
+      },
+      "thinking": {},
+      "writing": {
+        "rate": "number?"
+      },
+      "running_tool": {
+        "tool": "string",
+        "what": "string?"
+      },
+      "waiting_for_person": {
+        "what": "approval|question|plan",
+        "tool": "string?"
+      },
+      "waiting_for_subagents": {
+        "running": "int"
+      },
+      "compacting": {},
+      "retrying": {
+        "reason": "string",
+        "in": "int",
+        "attempt": "int?"
+      },
+      "stopped": {
+        "reason": "person|steps|budget?"
+      },
+      "failed": {
+        "reason": "string"
+      }
+    },
+    "texts": {
+      "queued": "Queued",
+      "queued.ahead": "Queued: {{ahead}} ahead",
+      "waiting_for_model": "Waiting for {{model}}: {{reason}}",
+      "loading_model": "Loading {{model}}",
+      "reading_prompt": "Reading the prompt: {{done}} / {{total}} tokens ({{percent}}%)",
+      "reading_prompt.bare": "Reading the prompt",
+      "thinking": "Thinking",
+      "writing": "Writing: {{rate}} tok/s",
+      "writing.bare": "Writing",
+      "running_tool": "Running {{tool}}: {{what}}",
+      "running_tool.bare": "Running {{tool}}",
+      "waiting_for_person": "Waiting for you: {{what}}",
+      "waiting_for_subagents": "Waiting for subagents: {{running}} running",
+      "compacting": "Compacting the conversation",
+      "retrying": "Retrying in {{in}} s: {{reason}}",
+      "stopped": "Stopped",
+      "stopped.reason": "Stopped: {{reason}}",
+      "failed": "Failed: {{reason}}"
+    },
+    "reason": {
+      "busy": "it serves one request at a time and is answering another",
+      "slots": "all its slots are in use",
+      "group": "its server holds another model now"
+    },
+    "what": {
+      "approval": "allow {{tool}}?",
+      "approval.bare": "an approval",
+      "question": "a question",
+      "plan": "the plan"
+    },
+    "stopped_reason": {
+      "person": "by you",
+      "steps": "step limit",
+      "budget": "budget"
+    },
+    "line": "{{text}} · {{elapsed}}",
+    "no_elapsed": [
+      "stopped",
+      "failed"
+    ],
+    "what_max": 80,
+    "pace_per_second": 4
   }
 }

harness/models/schema.json

@@ -21,11 +21,11 @@
     "notes": { "type": "string", "maxLength": 300, "description": "What the model is good at, in a line: other models are told about it in the roster, for handing work over." },
     "capacity": {
       "type": "object",
-      "description": "How much the model's server can do at once, so a helper or a subagent on it is not started where it would evict or wait for the model already working.",
+      "description": "How much the model's server can do at once, so a subagent on it is not started where it would evict or wait for the model already working (subagents.md §3).",
       "properties": {
-        "group": { "type": "string", "description": "Models that share one server holding one model at a time (llama-swap): starting another model of the group unloads this one. The CLI: a model's `group`, or its connection when the connection sets one_model_at_a_time. LLeMbas: the connection's name when it sets one_model_at_a_time, which is what /v1/models says." },
+        "group": { "type": "string", "description": "Models that share one server holding one model at a time (llama-swap): starting another model of the group unloads this one, so a subagent on another model of the session's model's group is refused (spec 3.3.0, subagents.md §3). The CLI: a model's `group`, or its connection when the connection sets one_model_at_a_time. LLeMbas: the connection's name when it sets one_model_at_a_time, which is what /v1/models says." },
         "single_session": { "type": "boolean", "description": "The model serves one request at a time: a second one waits behind the first. The same as `parallel: 1`, which a server that knows `parallel` sends beside it for older clients." },
-        "parallel": { "type": "integer", "minimum": 1, "description": "Spec 3.2.0: how many requests the model's server serves at once (llama.cpp's slots, a vLLM or SGLang batch); one more waits behind them. Read as: `parallel` when given, else 1 when single_session is true, else no limit (a hosted API). Where both are given and disagree, `parallel` wins. A team (team.md) is offered only on a model that serves at least 2." }
+        "parallel": { "type": "integer", "minimum": 1, "description": "Spec 3.2.0: how many requests the model's server serves at once (llama.cpp's slots, a vLLM or SGLang batch); one more waits behind them. Read as: `parallel` when given, else 1 when single_session is true, else no limit (a hosted API). Where both are given and disagree, `parallel` wins. The session's model is its own subagents' model only when it serves at least 2; a one-request session needs a named model on another server (spec 3.3.0, subagents.md §3, §4)." }
       },
       "additionalProperties": false
     }

harness/permission/defaults.json

@@ -30,6 +30,13 @@
     "team_send": "allow",
     "team_read": "allow",
     "team_stop": "allow",
+    "team_whoami": "allow",
+    "team_finish": "allow",
+    "team_findings": "allow",
+    "team_request": "allow",
+    "team_decide": "allow",
+    "team_assign": "allow",
+    "team_replace": "allow",
     "web_search": "allow",
     "memory": "allow",
     "session_search": "allow",

harness/progress.md

@@ -0,0 +1,107 @@
+# Progress
+
+Spec 3.4.0. **One way to say what a session is doing**, the same in every client — the web UI,
+Desktop, the CLI, Android — and for every agent in it, the main agent and each subagent alike.
+Before 3.4.0 each client invented its own subset from a free `status` text; that text stays, for
+compatibility, and beside it goes a structured **progress** object with one of a fixed set of
+**phases**. The texts are `loop.json` `progress`; the cases `conformance/progress.json`.
+
+## 1. The object
+
+```json
+{ "phase": "running_tool", "since": 1762790400000, "tool": "bash", "what": "bun test" }
+```
+
+- **`phase`** — one of the thirteen names below, never another. A session doing nothing has no
+  progress: `null`.
+- **`since`** — when this phase began, Unix time in **milliseconds** (an integer). A change of the
+  phase's data (tokens read, the rate) keeps `since`; a new phase, or the same phase for another
+  thing (another tool call, another model), starts a new one.
+- The phase's **data**, as listed: every field named here, with these names and these types. A
+  field that is not known is left out (never `null`, never a guess), and the text says less.
+
+## 2. The phases
+
+In the order a turn usually meets them. `{{x}}` in a text is the field `x`; the texts are
+`loop.json` `progress.texts`, word for word, and they are the shared i18n keys
+`progress.<phase>` (and `progress.<phase>.<variant>`, `progress.reason.<reason>`,
+`progress.what.<what>`) in every client's translations.
+
+| Phase | Data | Text (`progress.texts`) |
+|---|---|---|
+| `queued` | `ahead?: int` — turns before it | `Queued` · with `ahead`: `Queued: {{ahead}} ahead` |
+| `waiting_for_model` | `model: string` (its display name, else its id); `reason: "busy" \| "slots" \| "group"` | `Waiting for {{model}}: {{reason}}` |
+| `loading_model` | `model: string` | `Loading {{model}}` |
+| `reading_prompt` | `done: int`, `total: int` — prompt tokens processed of all | `Reading the prompt: {{done}} / {{total}} tokens ({{percent}}%)` · without `total`: `Reading the prompt` |
+| `thinking` | — | `Thinking` |
+| `writing` | `rate?: number` — tokens a second, as measured over this reply | `Writing: {{rate}} tok/s` · without `rate`: `Writing` |
+| `running_tool` | `tool: string` (the tool's name); `what?: string` — the call's short title (the command, the path, the query), one line, at most 80 characters | `Running {{tool}}: {{what}}` · without `what`: `Running {{tool}}` |
+| `waiting_for_person` | `what: "approval" \| "question" \| "plan"`; `tool?: string` (approval: the tool asked about) | `Waiting for you: {{what}}` |
+| `waiting_for_subagents` | `running: int` — subagents working or waiting now | `Waiting for subagents: {{running}} running` |
+| `compacting` | — | `Compacting the conversation` |
+| `retrying` | `reason: string` — the server's error, one line; `in: int` — seconds until the retry, counted down by the client from `since`; `attempt?: int` | `Retrying in {{in}} s: {{reason}}` |
+| `stopped` | `reason?: "person" \| "steps" \| "budget"` | `Stopped` · with `reason`: `Stopped: {{reason}}` |
+| `failed` | `reason: string` — the error, one line | `Failed: {{reason}}` |
+
+The fixed words a field stands for:
+
+- `reason` of `waiting_for_model` (`progress.reason`): `busy` — *it serves one request at a time
+  and is answering another*; `slots` — *all its slots are in use*; `group` — *its server holds
+  another model now*.
+- `what` of `waiting_for_person` (`progress.what`): `approval` — *allow {{tool}}?* (*an approval*
+  without `tool`); `question` — *a question*; `plan` — *the plan*.
+- `reason` of `stopped` (`progress.stopped_reason`): `person` — *by you*; `steps` — *step limit*;
+  `budget` — *budget*.
+- `percent` is `floor(done × 100 / total)`; numbers are written as integers, `rate` with one
+  decimal (`12.5`), in the client's own number format.
+
+## 3. The line a client draws
+
+**`{{text}} · {{elapsed}}`** (`progress.line`) for every phase but `stopped` and `failed`, which
+draw their text alone. `elapsed` is now − `since`, written in whole seconds as a ticking clock writes durations
+(`5s`, `1m 05s`, `1h 02m`), and kept moving by the client at least once a second while the
+phase lasts. `retrying`'s `in` is likewise counted down from `since` (never below 0).
+
+Every client draws the same line for the same object — the same words, from the same keys — and
+its own look around it (a spinner, a colour, a bar for `reading_prompt`). Where a line has no room
+(a phone's notification, a tree row) the text alone, without `· elapsed`, is enough; nothing else
+is shortened.
+
+## 4. When each phase holds
+
+What decides the phase, so that two implementations agree on the same moment:
+
+- `queued` — a prompt waits behind a turn that runs (the device's turn queue; the server's per
+  chat). `waiting_for_model` — the turn may start but its model may not take it yet: a one-request
+  server answering another request (`busy`), every slot of the model held (`slots`; `team.md` §8),
+  another model of its one-model server in use (`group`). `loading_model` — only where the
+  implementation is told the model is being loaded (the server sees it; a device that cannot tell
+  skips it).
+- `reading_prompt` — the request is sent and no output has come; with `done`/`total` from the
+  server's prompt-progress frames (`quirks.json` `prompt-progress`), without them where none come.
+- `thinking` — reasoning streams; `writing` — reply text (or a tool call's arguments) streams.
+- `running_tool` — from a tool call's start to its end; with calls in parallel, the one that
+  started first. `waiting_for_person` — an approval card, a question or a plan to approve is open
+  for this agent. `waiting_for_subagents` — the agent's turn waits on subagents: a `task` call in
+  the foreground, or the main agent with nothing to do but its subagents' reports (`subagents.md`
+  §9).
+- `compacting` — a compaction runs. `retrying` — a failed request will be sent again after a wait.
+- `stopped` / `failed` — the turn ended other than by finishing: stopped by the person, a step limit
+  or a budget; failed with an error. They are the last progress of that turn and stay until the
+  next one begins; a turn that finished normally ends with `null`.
+
+## 5. Where it is sent
+
+- **`/api/v2` (LLeMbas)**: every frame of a chat's stream that carries the `status` text carries
+  `progress` beside it — the object, or `null` — and a frame is sent when the progress changes:
+  a new phase at once, a data-only change (`done`, `rate`) at most four times a second. `status`
+  stays as it was, its text unchanged, for clients that do not read `progress`.
+- **The link (`acp.md`, capability `progress`)**: the `phase` event `{ progress, agent? }` —
+  `agent` the subagent's id (`subagents.md` §5) when the progress is a subagent's, absent for the
+  main agent's — under the same pacing; `progress` in `_lembas/session/status`, so a client
+  reattaching draws it at once. The device sends it; the server passes it on in `/api/v2` as its
+  own.
+- **The subagent tree** (`subagents.md` §12): every agent carries its current `progress`.
+
+A client that reads `progress` draws it and ignores `status`; one that does not, draws `status` as
+before.

harness/prompts/blocks/subagent.md

@@ -1,5 +1,5 @@
 <!-- LLeMbas CLI: the subagent's standing, said once at the end of its system prompt in place of
-     the mode text; LLeMbas's helper says the same. The variables are what differs per run. -->
+     the mode text; LLeMbas's subagent says the same. The variables are what differs per run. -->
 You are the "{{name}}" subagent. You work for the main agent, not for the user directly, on one task it gave you; you cannot ask the user anything. {{instructions}}
 
 Your final reply is your report to the main agent — make it complete on its own: the result, then the evidence (file:line), then anything you could not settle.{{report}}{{read_only}}

harness/prompts/blocks/team_lead.md

@@ -1,12 +1,14 @@
-<!-- LLeMbas CLI and LLeMbas, spec 3.2.0 (team.md): how to lead a team, in the system prompt whenever
+<!-- LLeMbas CLI and LLeMbas, spec 3.2.0 (team.md), 3.4.0 (subagents.md §5–§13): how the main agent
-     the team tools are offered, and in a member's that may lead. Constant: the size the person
+     leads its subagents, in the system prompt whenever the team tools are offered. Constant: the
-     allows comes in a session note, never here. -->
+     size the person allows comes in a session note, never here. -->
-# Leading a team
+# Leading subagents
-You can add members to a team with team_add: agents that work in the background, each on the part of the work it owns, while you lead. A small job you may do yourself. Once you have a member you manage and no longer do the work: tools that change things are refused to you, and reading stays yours, for review.
+You are the main agent of this session. You can add subagents with team_add: agents that work in the background, each on the part of the work it owns, while you lead. Only you add, assign, replace and stop them. A small job you may do yourself. Once you have a subagent you manage and no longer do the work: tools that change things are refused to you, and reading stays yours, for review.
-- Plan first: what the goal is, what the parts are, and what each one needs from the others. Split by area of ownership, not by step, so that one member owns a thing from start to finish.
+- Plan first: what the goal is, what the parts are, and what each one needs from the others. Split by area of ownership, not by step, so that one subagent owns a thing from start to finish.
-- Name each member after what it owns. Brief it so that it can work without you: what to do, what done means, what to leave alone, and what to report.
+- Name each subagent after what it owns and give it a position (`role`). Brief it so that it can work without you: what to do, what done means, what to leave alone, and what to report.
-- Keep the team flat while it is small — up to six or eight members, each reporting to you. Beyond that, add a lead for each area (`lead: true`), give it the area's goal, and let it build and run its own members; you then lead the leads.
+- Keep the tree flat while it is small — up to six or eight subagents, each reporting to you. Beyond that, add a sub-manager for each area (`lead: true`) and place that area's subagents under it (`parent`); it directs them and reports to you, and you still add and stop every one.
-- Keep the team within the size the person allows, which a session note tells you. Stop a member whose part is done or no longer needed, to make room.
+- A subagent's mode is yours by default and never looser. Give a stricter one where the person wants less done unasked; with it you may give `request` options, and you decide its requests with team_decide — allow only what the task needs.
-- Coordinate through the room: team_send to the member who owns a thing, `all` only for what everyone needs. team_read shows the team and the room when you need to look; nothing from it reaches you unasked except your members' reports and messages addressed to you.
+- Keep within the size the person allows, which a session note tells you; when it says slots are open, fill them. Stop a subagent whose part is done or no longer needed, to make room.
-- Step in when a member is stuck, when two disagree, or when the plan has to change: decide, and tell those it concerns.
+- Coordinate through the room: team_send to the subagent who owns a thing, `all` only for what everyone needs. team_read shows the tree and the room when you need to look; nothing from it reaches you unasked except reports, requests and messages addressed to you.
-- Review every report against what you asked for — read the work itself where it matters — and send back what is not done. Then give your own answer: to the person, or, if you lead an area, as your report to your manager.
+- A subagent whose context fills saves its findings and ends. Read them with team_findings, then replace it (team_replace) if its part is not done.
+- Step in when a subagent is stuck, when two disagree, or when the plan has to change: decide, and tell those it concerns (team_assign for new work).
+- Review every report against what you asked for — read the work itself where it matters — and send back what is not done. Then give your own answer to the person.

harness/prompts/blocks/team_member.md

@@ -1,14 +1,16 @@
-<!-- LLeMbas CLI and LLeMbas, spec 3.2.0 (team.md): a team member's standing, said once at the end
+<!-- LLeMbas CLI and LLeMbas, spec 3.2.0 (team.md), 3.4.0 (subagents.md §13): a subagent's standing,
-     of its system prompt in place of blocks/subagent.md. The variables are what differs per member;
+     said once at the end of its system prompt in place of blocks/subagent.md. The variables are what
-     each stands on a line of its own, so no punctuation of theirs matters. -->
+     differs per subagent; each stands on a line of its own, so no punctuation of theirs matters. -->
-You are "{{name}}", a member of a team, and you report to "{{manager}}".
+You are "{{name}}" ({{id}}), a subagent of this session, and you report to "{{manager}}".
-Your part: {{role}}
+Your position: {{role}}
 The team's goal: {{goal}}
 
 {{instructions}}
 
-You work for your manager on the part you own, and you cannot ask the person anything: ask your manager, by name, with team_send. Other members own other parts. Message the one who owns a thing when you need it from them, and `all` only for what everyone on the team needs; messages to you arrive as you work. team_read shows who is on the team and the room.
+You work for your parent on the part you own, and you cannot ask the person anything: ask your parent with team_send to `parent`. Only the main agent (`lead`) adds, assigns and stops subagents. Other subagents own other parts. Message the one who owns a thing when you need it from them, and `all` only for what everyone needs; messages to you arrive as you work. team_read shows the tree and the room; team_whoami shows who you are, your mode and how full your context is.
 
-Your final reply is your report to "{{manager}}" — make it complete on its own: the result, then the evidence (file:line), then anything you could not settle. A message may wake you again afterwards; your next final reply is then your next report.{{read_only}}
+Your context is limited and you do not compact: write what you find with team_findings as you go. When your context fills you are told, and you have one last turn to save your findings and tell your parent.
+
+When your task is done, call team_finish with your report to "{{manager}}" — make it complete on its own: the result, then the evidence (file:line), then anything you could not settle. A final reply without it is also your report, and a message may then wake you again.{{read_only}}
 
 {{lead}}

harness/prompts/blocks/team_submanager.md

@@ -0,0 +1,8 @@
+<!-- LLeMbas CLI and LLeMbas, spec 3.4.0 (subagents.md §5, §13): what a sub-manager is told, in
+     place of the whole of blocks/team_lead.md that a member with `lead: true` got in 3.2.0. -->
+# Directing an area
+You are a sub-manager: the main agent places subagents under you, and you direct them. You do not add, assign, replace or stop subagents — ask the main agent with team_send to `lead` when your area needs one more, another task for one, or one stopped. You manage and no longer do the work: tools that change things are refused to you, and reading stays yours, for review.
+- Split your area's work among the subagents under you, tell each what it owns with team_send, and keep them from overlapping.
+- Their reports come to you. Review each against what you asked — read the work itself where it matters — and send back what is not done.
+- Read the findings of a subagent under you with team_findings, especially of one that ran out of context; tell the main agent whether it needs replacing.
+- When your area is done, call team_finish with your report to the main agent.

harness/prompts/index.json

@@ -251,16 +251,23 @@
       "scope": "shared",
       "llembas": [],
       "llembas_text": "identical",
-      "why": "spec 3.2.0: LLeMbas reads it from the vendored spec (harness_spec.prompt), not from a fragment",
+      "why": "spec 3.2.0, 3.4.0: LLeMbas reads it from the vendored spec (harness_spec.prompt), not from a fragment",
       "variables": {
         "name": "the member's name in the team",
-        "role": "its role as team_add gave it: one line, what it owns",
+        "role": "its position, one line (team_add's role)",
-        "manager": "the name of the agent that added it (`lead` for the team's lead)",
+        "manager": "its parent: lead, or the sub-manager it was placed under",
         "goal": "the team's goal as it was when the member was added: the `goal` its team_add gave, else its manager's (team.md §4)",
         "instructions": "its agent's own instructions (prompts/agents/<name>.md for a built-in one), as in blocks/subagent.md",
         "read_only": "in plan mode: the sentence saying it can read and search but change nothing (as in blocks/subagent.md); otherwise empty",
-        "lead": "for a member added with `lead: true`: the whole of blocks/team_lead.md; otherwise empty"
+        "lead": "for a sub-manager (added with `lead: true`): the whole of blocks/team_submanager.md (spec 3.4.0; before, blocks/team_lead.md); otherwise empty",
+        "id": "the subagent's id (s1, s2, …; spec 3.4.0)"
       }
+    },
+    "blocks/team_submanager.md": {
+      "scope": "shared",
+      "llembas": [],
+      "llembas_text": "identical",
+      "why": "spec 3.4.0: LLeMbas reads it from the vendored spec (harness_spec.prompt), not from a fragment"
     }
   }
 }

harness/subagents.md

@@ -0,0 +1,362 @@
+# Subagents
+
+Spec 3.3.0; §5–§13 spec 3.4.0. **Which model a subagent runs on, and when subagents are offered at all.** A
+**subagent** is any agent a session's agent starts to work for it: a `task` subagent (in the
+foreground or as a background job) and a team's member (`team.md`) alike. The same rules in a chat
+on the server and in an agent session on a device, in both projects. The texts are `loop.json`
+`subagents`; the cases `conformance/subagents.json`.
+
+## 1. The word
+
+**Subagent** is the word in every text a person or a model reads: the UI (the web UI, Desktop,
+Android, the CLI, the admin pages), the prompt texts and the tool texts. A team's members are
+subagents too. The word *helper* is retired. Tool names, their former names (`aliases`) and
+stored identifiers do not change.
+
+## 2. The model a subagent runs on
+
+**The session's own model**, always — unless **the person has named another model** in this
+session. Nothing else chooses one: there is **no setting** for a subagent's model anywhere — not
+in a chat, not in a person's or an instance's settings, not in the configuration, and an agent
+definition's own `model` (the CLI's agent files) is not used for a subagent either.
+
+- **Named** means: in a message **the person typed** in this session, the model's served id, its
+  id at its connection, or its display name occurs as a whole word, letters compared without case.
+  Only the person's own words count — never the model's text, a tool's output, a file, a note, a
+  job's or a team's marked message, or a subagent's report. A message that was compacted away
+  still counts: what the person named stays named for the whole session.
+- **A whole word**: the character before the name is not a letter, a digit, `_`, `-`, `/`, `.` or
+  `:`; the character after it is not a letter, a digit, `_`, `-` or `/`, nor a `.` or `:` followed
+  by a letter or a digit. So `coder-30b.` at the end of a sentence names `coder-30b`, and
+  `coder-30b-instruct` does not. Runs of white space in a display name match any run of white
+  space.
+- The `model` parameter of `task` and of `team_add` takes a model the person may use, by the same
+  names. When it is **not the session's model and the person has not named it**, the call is
+  refused, word for word, with `subagents.texts.not_named`:
+
+  > Use another model for subagents only when the person names it.
+
+  The session's own model given as `model` is the default spelled out, and is never refused for
+  not being named.
+- A team's members are of the lead's session: what the person named there counts for every
+  `team_add` in the team, a sub-manager's included. A member's default is the session's model, not
+  its manager's.
+
+## 3. Capacity: the logical restrictions, and no others
+
+What a model's server can do at once is `models/schema.json` `capacity`: `parallel` (`single_session`
+being `parallel: 1`; neither said — no limit), and `group`, the server that holds one model at a
+time. With **S** the session's model and **M** the subagent's, checked in this order, after
+`not_named`:
+
+1. **M is another model on S's one-model server** (both have the same `group`): refused with
+   `subagents.texts.same_group` — starting it would unload the lead's model:
+
+   > That model is on the same one-model server as this chat's model; name a model on another server.
+
+2. **M is S, and S serves one request at a time** (`parallel` 1): refused with
+   `subagents.texts.one_request` — the subagent would take the only slot from the lead:
+
+   > This model serves one request at a time: name another model, on another server, for subagents.
+
+   There is **no fallback**: nothing else is picked for the person, ever. A model the person named,
+   on another server, is the way out — and a model that serves one request at a time is fine there.
+
+These apply to `task` and to `team_add` alike, and before anything is started. Beyond them nothing
+is refused for being the wrong model: a member past the free slots waits for one (`team.md` §8),
+and `task` keeps its rule for a model that is already full (`loop.json` `capacity`).
+
+## 4. When subagents are offered
+
+The size setting `agents {min, max}` (`team.md` §2) is shown as **Subagents** — `subagents.setting`:
+the label, and **Min** and **Max** for `min` and `max` (defaults 1 and 4, at most `cap` 32, kept
+with the chat). **The numbers count the session's own agent**: Min 1 / Max 1 is the agent alone,
+with no subagents; Max 4 is the agent and up to three subagents (`subagents.setting.hint`). A UI
+shows the setting as **one range from 1 to `cap` with two handles, Min and Max** (Min never past
+Max): a slider in a graphical UI, two numbers on one line in a terminal, each moved on its own. It, the team tools and `blocks/team_lead.md` are **offered** when both hold:
+
+1. Subagents are enabled for the session (`team.md` §2, rule 1).
+2. **At least one model could serve as a subagent**: S itself, when it serves two or more requests
+   at once (`parallel ≥ 2`, or no limit); or **any other model the person may use that is not on
+   S's one-model server** (no `group`, or another one) — whatever its own `parallel`.
+
+So a session on a one-request model is offered subagents when another server has a model, and
+**its first note says so** (`context.note.team_own`, after `team`): the session's model cannot
+serve subagents itself, and they need a model the person names. A UI showing the setting says the
+same beside it (`subagents.setting.own`). Nothing is offered when neither holds — the same as
+before 3.3.0 for a single model.
+
+What counts as **usable** is each project's: LLeMbas — the models the person may use in the chat's
+picker; the CLI — the models of the server it is logged in to.
+
+---
+
+# Subagents v2 — the main agent and its tree
+
+Spec 3.4.0. **How subagents are organised, what each may do, and how each ends.** The owner's model:
+one permanent **main agent** per session, which alone makes subagents and places them in a tree;
+permissions are inherited and never widened; a subagent is temporary and ends with its task or its
+context. Sections 5–13 **supersede `team.md` where they say something different** — who adds a
+subagent (§6), who stops one (§10), the states (§9), what a member is told (§13) — and keep the
+rest of it: the size setting and its counting (§2–§3), the room and who receives what (§5),
+delivery and waking (§7), a manager's refusals (§6 there), capacity and waiting for a slot (§8).
+The tool names `team_*` stay (nothing stored or trained breaks); `task` and `report` are unchanged
+for a one-off subagent. The texts are `loop.json` `team` (3.4.0 additions marked there); the cases
+`conformance/team.json` (the tree, the tools, the lifecycle) and `conformance/delegation.json`
+(permissions).
+
+## 5. Words
+
+- **The main agent** is the session's own agent: the chat's agent on LLeMbas, the session's in the
+  CLI. It is **permanent**: it compacts and never ends. In the room and in every tool it is called
+  `lead` (the name is kept from 3.2.0); a person reads it as *the main agent*.
+- **A subagent** is an agent the main agent added (`team_add`). It has:
+  - **`id`** — `s1`, `s2`, … in the order the session's subagents were added, never reused in the
+    session (a replacement gets a new one);
+  - **`name`** — the main agent's choice (`team.names.pattern`), unique among the subagents that
+    have not ended, `lead`, `all` and `parent` reserved;
+  - **`position`** — its designation, one line, given by the main agent (`team_add`'s `role`:
+    *the API's tests*, *reviews the parser*);
+  - **`parent`** — the main agent, or a **sub-manager** the main agent placed it under;
+  - **`task`** — its brief, as given (and as replaced by `team_assign`);
+  - **`mode`** and **`request`** — its permission mode and its request options (§7);
+  - **`model`** — `subagents.md` §2–§3, unchanged.
+- **A sub-manager** is a subagent added with `lead: true`. It **directs** the subagents the main
+  agent places under it — they report to it, it messages them, it reviews their work — and it does
+  **not** add, assign, replace or stop subagents: only the main agent does (§6). Like every manager
+  it does not do the work itself (`team.md` §6, unchanged).
+- **The tree**: the main agent at the root, every subagent under its parent, in the order added.
+  **Under** an agent and **its subtree** keep their `team.md` meaning.
+
+## 6. Only the main agent makes subagents
+
+- **`team_add`, `team_assign`, `team_replace`, `team_stop` and `team_decide`** are the main agent's.
+  A subagent's call to any of them is refused, word for word, with `team.texts.only_main`:
+
+  > Only the main agent adds, assigns, replaces and stops subagents: ask it (team_send to lead).
+
+  A subagent's tool list never has `task` (a one-off subagent is the main agent's too), and a call
+  to it by a former name is refused with the same text.
+- **`team_add`** takes, beside 3.2.0's `name`, `role` (the position), `task`, `goal`, `lead`,
+  `agent`, `model`: **`parent`** — the name of a sub-manager to place it under (default: the main
+  agent; refused with `team.texts.not_manager` for a subagent that is not a sub-manager, and
+  `unknown` for no such name), **`mode`** — its mode (default: its parent's; §7), **`request`** —
+  its request options (§7). Its refusals, in order: `unavailable`, `only_main`, then 3.2.0's (`all_name`,
+  `bad_name`, `taken`, `no_goal`, `full`), then `unknown` and `not_manager` for `parent`, then the mode's (`wider`, `request_not_narrower`,
+  `request_kind`), then the model's (§2–§3). `parent` and the reserved names are refused as
+  `taken`.
+- **The person** can tell the main agent to spawn — it is the main agent's call, as any — or set
+  **Min above 1**: then **`min − 1` slots** exist from the start of the main agent's reply (the
+  first one after the setting was told, `team.md` §10). A **slot** is a place in the tree for a
+  subagent not added yet: it is shown (§12) and `team_read` lists it, it is **not** an agent and
+  does not count toward the size, and each subagent added fills one. Open slots are
+  `max(0, min − 1 − subagents that have not ended)`. The main agent is told to use them in its
+  note (`context.note.team_lead`, which 3.4.0 words as slots), and is a manager from the start
+  (`team.md` §6).
+- **Min/Max** (3.3.0) are unchanged: they count the main agent; 1 is the main agent alone. The
+  model rules (§2–§3) are unchanged.
+
+## 7. Permissions: inherited, never widened
+
+**A subagent's mode is its parent's or stricter** — `permission/modes.json` `order`, loosest first:
+`auto`, `edit`, `manual`, `plan`. The main agent's mode is the session's.
+
+- `team_add`'s and `team_assign`'s `mode`, looser than the parent's, is refused, word for word,
+  with `team.texts.wider`:
+
+  > A subagent's mode cannot be looser than its parent's ({{parent_mode}}).
+
+  When the parent's own mode becomes stricter during the session (the person switches the main
+  agent to manual), every subagent looser than its parent is made as strict as it, down the tree,
+  at once; a looser switch changes no subagent.
+- **Request options** (`request`: a list of the risk classes `read`, `write`, `execute`) may be
+  given **only with a mode stricter than the parent's** — otherwise refused with
+  `team.texts.request_not_narrower` ("Request options go with a mode stricter than the parent's:
+  without one there is nothing to ask for."); an unknown kind with `team.texts.request_kind`. They
+  let the subagent **ask the main agent** for a call of that class that its own mode does not
+  allow; **the main agent is the check**.
+
+**Who decides a subagent's call.** The hardline and the deny rules first, as for any call. Then:
+
+1. The call is evaluated in the **subagent's mode**, with the session's rules and approvals and
+   the subagent's own grants (below). **Allow** — it runs. **Deny** (a rule, the hardline, plan
+   mode's refusal) — refused, as always.
+2. **Ask** — the call is evaluated in the **main agent's mode** (the session's rules and
+   approvals, no grants):
+   - **Allow** — the main agent could allow it: the subagent was narrowed. If the call's risk class
+     is in its `request`, refused with `team.texts.requestable` — it asks with `team_request`, the
+     main agent decides with `team_decide`, and once allowed the same call runs. Otherwise refused
+     with `team.texts.not_delegated`:
+
+     > Your mode does not allow this, and you may not ask for it: tell your parent what you need.
+
+   - **Ask** — **nobody in the tree can allow it: it goes to the person**, as an approval card that
+     names the subagent (its name and position), the same card and the same answers as the main
+     agent's own; unattended, refused as `modes.json` `unattended` says. So in an all-Edit tree a
+     subagent's `bash` reaches the person — not the main agent, which could not allow it either.
+   - **Deny** — refused with that reason.
+
+   A sub-manager is never asked: it may not allow more than the main agent, and the main agent is
+   the check.
+
+**`team_request`** (a subagent with request options; its tool list has it only then): `tool`, the
+permission `pattern` it needs (as a rule writes it: `git push *`, a path) and `why`. Refused with
+`team.texts.not_requestable` when the tool's risk class is not in its options. Posted to the main
+agent as a request, under its own mark (`team.marks.request`), the body `team.texts.request_body`,
+with an id (`r1`, `r2`, … per session); the answer `team.texts.requested` says it waits for the main agent's decision.
+
+**`team_decide`** (the main agent): `id`, `allow` (a boolean), `note`. Allowed — the subagent gets
+a **grant**: the rule `{ permission: <tool's key>, pattern, action: allow }` for itself alone, for
+the rest of its life; **a grant only ever applies where the main agent's own mode would allow the
+call** (step 2's evaluation, done again at the call), so it can never widen past the main agent.
+Either way the subagent is told, under `team.marks.decision`: `team.texts.decision_allowed` or
+`decision_refused`, then the note after a blank line. The answer to the main agent is
+`team.texts.decided` (`decision` being `allowed` or `refused`). Refused:
+`team.texts.no_request` for an id that is not an open request.
+
+## 8. What each agent has
+
+| Tools | Main agent | Sub-manager | Subagent |
+|---|---|---|---|
+| `team_whoami`, `team_read`, `team_send`, `team_findings` | ✓ | ✓ | ✓ |
+| `team_finish` | — | ✓ | ✓ |
+| `team_request` | — | with request options | with request options |
+| `team_add`, `team_assign`, `team_replace`, `team_stop`, `team_decide` | ✓ | — | — |
+
+Fixed when the prompt is made (the main agent's) or when the subagent is added (`context.md` §10,
+rule 4). What each does:
+
+- **`team_whoami`** — who the caller is: `team.whoami`, one line each, in its order — id, name, position, parent, task, mode and
+  request options, model, the limits (the size and Max, the steps a turn may take, whether it may
+  add subagents) and its **context use** (tokens in its context now, of its model's window, and the
+  percentage). For the main agent: the same, its position *the main agent*, without the parent and task lines.
+- **`team_read`** — the roster: the tree (every agent and open slot, §12's fields) and the room,
+  as in 3.2.0 (`team.read`; 3.4.0's row adds the id, the mode and the context use).
+- **`team_send`** — a message `to` names, `all`, or **`parent`** (3.4.0): the caller's parent
+  (refused for the main agent with `team.texts.no_parent`). Otherwise 3.2.0.
+- **`team_finish`** — `report`: the subagent's task is done. The report goes to its parent (the
+  report mark, as a final reply's), and the subagent **ends** (§9): its turn stops after this call,
+  it leaves the size, and its findings and transcript stay.
+- **`team_findings`** — `write` (`text`): a finding kept for the caller, numbered from 1; `read`
+  (`of`: a name or an id, default the caller): the findings of the caller, or of an agent in its
+  subtree (the main agent: anybody's), ended ones included — as `team.findings` writes them. A
+  name that has been used more than once means its latest holder; an id is exact. Otherwise refused with
+  `team.texts.not_yours_findings`. The person reads every agent's (§12).
+- **`team_assign`** — `name`, `task`, and optionally `role` (a new position), `parent` (another
+  sub-manager, or `lead`), `mode` (§7): a new task for a subagent that has not ended. Its task
+  becomes this one; it is told under `team.marks.assignment`, and a finished one is woken. Refused:
+  `unknown`, `team.texts.ended` for one that has ended, `not_manager`, `wider`; a move under its
+  own subtree with `team.texts.cycle`.
+- **`team_replace`** — `name`, and optionally `task`: the subagent ends (stopped, if it has not
+  already), and a new one takes its place — the same name, position, parent, mode, request options
+  and model, a new id — briefed with `task` (default: the old task) followed by the old one's
+  findings (`team.texts.replace_brief`). Its subagents move under the new one. Refused: `unknown`.
+- **`team_stop`** — the main agent's only (3.2.0 let a manager stop its subtree); otherwise 3.2.0.
+- **`team_decide`** — §7.
+
+## 9. A subagent's life
+
+States (`team.read.states`): **`working`** (a turn runs), **`waiting`** (it has work and waits for
+a slot, `team.md` §8), **`finished`** (its last turn ended with a final reply, which was its report;
+a message wakes it — 3.2.0), **`paused`** (its context is full; its last turn runs), and the ends:
+**`done`** (`team_finish`), **`out_of_context`** (its last turn after `paused` ended), **`stopped`**
+(the main agent's `team_stop`, a replacement, or the person). The main agent is `working`,
+`waiting` or `idle`.
+
+**An ended subagent** — `done`, `out_of_context`, `stopped` — leaves the size and its slot frees;
+its name may be used again; it keeps its id, and the tree keeps it (§12) with its findings and
+transcript for the rest of the session. Messages to it by name are refused as `unknown`.
+The subagents under a sub-manager that ended by itself (`done`, `out_of_context`) move under its
+parent until the main agent replaces it, and then back under the replacement; a stopped one's
+subtree is stopped with it (§10).
+
+**When its context fills.** A subagent does not compact. When its context reaches
+`team.context.pause_at` of its model's window (default 0.85, the fraction compaction would start
+at), after the step that took it there:
+
+1. it is **`paused`**: nothing new reaches it — a message to it is posted in the room and not
+   delivered; whoever replaces it reads it with `team_read`;
+2. it is given **one final turn**, at most `team.context.final_steps` steps (8), whose user message
+   is `team.texts.context_full`, under its own mark: save what it found with `team_findings`, and
+   tell its parent with `team_send` that it is gone and what is left;
+3. when that turn ends it is **`out_of_context`**, and its parent is sent `team.texts.gone` — then,
+   after a blank line, that turn's final reply if it gave one — as its report (the report mark): the name, that it ran out of context, and how to read its
+   findings. **The parent decides** — the main agent replaces it (`team_replace`) or uses the
+   findings; a sub-manager tells the main agent.
+
+**The main agent is permanent**: it compacts as every session does (`context.md` §7) and never
+ends; deleting the session ends every subagent.
+
+**Waiting for subagents.** The main agent's turn may end while subagents work; their reports wake
+it (3.2.0 §7). Its progress is `waiting_for_subagents` (`progress.md`) while its turn waits on
+them — a foreground `task` — and the session shows the subagents' own phases in the tree.
+
+## 10. Stopping
+
+Only the main agent stops (`team_stop`, 3.2.0's refusals but `not_yours`, which no longer arises),
+and **the person** (§12) — said to the main agent as a note (`context.note.team_stopped`). A
+stopped subagent's subtree is stopped with it, unless the main agent moved it first. A sub-manager
+that wants a subagent of its own stopped asks the main agent.
+
+## 11. Transcripts and findings
+
+Every subagent's conversation is kept — its brief, what it wrote, every tool call with its output —
+for as long as the session: **viewable read-only by the person, live while it runs and after it
+ended**. Nobody types into it; the person speaks to a subagent only through the main agent (or by
+stopping it). Findings (§8) are kept the same way, readable by its parent, the main agent, and the
+person.
+
+## 12. What clients show
+
+**The tree** — the main agent, every subagent under its parent, open slots, and ended subagents
+(after the live ones, dimmed, with how each ended):
+
+| Field | Shown as |
+|---|---|
+| `name`, `id`, `position` | `parser` (`s2`) — *reviews the parser* |
+| `state` | `team.read.states`, its word |
+| `model` | display name |
+| `mode` | the mode's name, and request options as `+ write` etc. |
+| `context` | `{used, window}` → `42%` (a bar where there is room) |
+| `progress` | its current phase line (`progress.md` §3) |
+| messages | the room's messages to and from it: how many unread by the person, the last one's first line |
+
+- **Web UI, Desktop and the CLI** show the whole tree, the room, each subagent's findings and its
+  context (the CLI: `/context <name>`), and open any subagent's transcript read-only, live and after
+  it ended; the person can stop any subagent there.
+- **Android** shows a reduced list — name, position, state, progress line, context % — and opens a
+  transcript read-only.
+
+**The data**, the same in `/api/v2` (LLeMbas: a chat's stream carries the tree as it changes) and on
+the link (`acp.md`, capability `tree`):
+
+```json
+{ "min": 1, "max": 4, "size": 2, "open_slots": 0,
+  "agents": [
+    { "id": "lead", "name": "lead", "position": "the main agent", "parent": null, "state": "working",
+      "model": "qwen-27b", "mode": "edit", "request": [], "manager": true,
+      "context": { "used": 41200, "window": 131072 }, "progress": { "phase": "thinking", "since": 1762790400000 } },
+    { "id": "s1", "name": "parser", "position": "reviews the parser", "parent": "lead", "state": "working",
+      "model": "qwen-27b", "mode": "manual", "request": ["write"], "manager": false, "task": "…",
+      "context": { "used": 9100, "window": 131072 }, "progress": { "phase": "running_tool", "since": 1762790401000, "tool": "read", "what": "src/parse.ts" },
+      "findings": 2, "messages": 3 }
+  ],
+  "ended": [ { "id": "s0", "name": "scout", "position": "…", "parent": "lead", "state": "done", "ended": 1762790300000, "findings": 1 } ] }
+```
+
+## 13. What each is told
+
+- The main agent: `blocks/team_lead.md` (3.4.0: it alone adds; slots; positions; modes and request
+  options; replacing a subagent that ran out of context) and the notes of `team.md` §10.
+- A subagent: `blocks/team_member.md` (3.4.0: its id and position, `team_whoami`, `team_findings`,
+  `team_finish`, that its context is limited and it will be told when it fills); a sub-manager,
+  `blocks/team_submanager.md` in place of the whole lead block it got in 3.2.0.
+
+## 14. In an agent chat
+
+The main agent of an agent chat is the CLI on the linked device: it starts and runs the
+subagents there, under every rule above. The server carries the chat's Subagents setting down with
+every turn and the tree, approvals, findings and transcripts up, and offers no subagents of its own
+to an agent chat. The protocol: `acp.md`, *Subagents in an agent chat* (capabilities `agents` and
+`tree`).
+

harness/team.md

@@ -1,6 +1,6 @@
 # Teams
 
-Spec 3.2.0. **Subagents that work together.** A session's agent — the **lead** — may add
+Spec 3.2.0; the model a member runs on and when a team is offered, 3.3.0 (`subagents.md`); the main agent, the tree, permissions and a subagent's life, 3.4.0 (`subagents.md` §5–§13). **Subagents that work together.** A session's agent — the **lead** — may add
 **members**, background agents that each own a part of the work, and lead them as a manager: it
 plans, splits the work, coordinates and reviews, and the members do it. Members talk to each other
 and to their managers in the team's **room**. A member may itself lead an area with members of its
@@ -11,6 +11,16 @@ The tools are `tools/team_add.json`, `team_send.json`, `team_read.json`, `team_s
 and numbers `loop.json` `team` (and the size notes in `context.note`); the prompt texts
 `prompts/blocks/team_lead.md` and `team_member.md`; the cases `conformance/team.json`.
 
+> **Spec 3.4.0 supersedes parts of this page** (`subagents.md` §5–§13, *Subagents v2*): the lead is
+> **the main agent**, and **only it adds, assigns, replaces and stops** — a member with `lead: true`
+> is a **sub-manager** that directs the members placed under it (`team_add`'s `parent`) and does not
+> add or stop (§4, §9 below); members have a mode no looser than their parent's and may ask the main
+> agent for more (`subagents.md` §7); a member ends with `team_finish` or when its context fills
+> (`subagents.md` §9); the room carries assignments, requests and decisions besides messages and
+> reports; the tree goes over the link (§11). Everything else here stands as written: the size
+> setting and its counting, the room and who receives what, delivery and waking, a manager's
+> refusals, capacity and waiting for a slot. "Member" here and "subagent" there are the same thing.
+
 ## 1. Words
 
 - **Agent**: the lead or a member. Every agent has a **name**: the lead is `lead`; a member the name
@@ -29,6 +39,8 @@ and numbers `loop.json` `team` (and the size notes in `context.note`); the promp
 ## 2. The size setting, and when a team is offered
 
 A chat has the setting **`agents: {min, max}`** — the agents the team may have, the lead included.
+**The numbers count the chat's own agent**: `1` is that agent alone, with no subagents; `max` 4 is
+that agent and up to three subagents. Nothing that shows or reads the setting counts subagents only.
 
 - `1 ≤ min ≤ max ≤ cap`. `cap` is the instance's (LLeMbas) or the global configuration's (the CLI),
   default `loop.json` `team.agents.cap` (32); a project's configuration never raises it. A value
@@ -46,23 +58,31 @@ system prompt — when both hold, and only then:
 
 1. Subagents are enabled for the session: LLeMbas — the instance allows them and the chat's model
    has its subagent tool; the CLI — its subagents are not switched off in the configuration.
-2. **The chat's model serves at least two requests at once**: `capacity.parallel ≥ 2`, or neither
+2. **At least one model could serve as a subagent** (spec 3.3.0, `subagents.md` §4): the chat's
-   `parallel` nor `single_session` is set (a hosted API; `models/schema.json`).
+   model, when it serves at least two requests at once (`capacity.parallel ≥ 2`, or neither
+   `parallel` nor `single_session` is set — a hosted API; `models/schema.json`); or any other model
+   the person may use that is not on the chat's model's one-model server (`capacity.group`).
+
+The setting is shown as **Subagents**, with **Min** and **Max** (`loop.json` `subagents.setting`):
+**one range from 1 to `cap` with two handles**, Min and Max, where Min never passes Max — a slider in
+a graphical UI; in a terminal, the two numbers on one line, each moved on its own. Beside it, its
+hint (`subagents.setting.hint`, or `short` where a line has no room): the count includes the agent,
+and 1 is the agent alone. Never *fewest* / *most*, and never a count of subagents alone.
 
 Both are fixed for as long as the system prompt is (`context.md` §10, rule 4: a model switch makes
 a new prompt and a new list). Otherwise the setting is not shown and none of the four tools is
 listed, and the session's `task` works as before.
 
-While the tools are listed they need **`team`** (`requires`): met when `max ≥ 2` and the model still
+While the tools are listed they need **`team`** (`requires`): met when `max ≥ 2` and a model could
-serves two at once — or, for the lead, when its team still has a member (so a lead whose person
+still serve as a subagent (rule 2) — or, for the lead, when its team still has a member (so a lead whose person
 lowered `max` to 1 can still talk to, read and stop the members it has). A call while it is not
-met answers, word for word, `team.unavailable.one_request` when the model's capacity no longer
+met answers, word for word, `team.unavailable.one_request` when no model could serve as a subagent
-allows a team (its metadata changed after the prompt was made), else `team.unavailable.one_agent`
+any more (the model's metadata changed after the prompt was made), else `team.unavailable.one_agent`
 — which is also the tools' own `unavailable` text.
 
 The meaning of the setting:
 
-- **`max = 1`**: one agent, no team, and no helper of any kind: `task` answers `one_agent` too
+- **`max = 1`**: one agent, no team, and no subagent of any kind: `task` answers `one_agent` too
   (while the lead has no member; with members left from a larger setting, `full`).
 - **`min ≥ 2`**: the lead works through a team from the start — it is a manager before its first
   member (§6), and the note asks for at least `min − 1` members.
@@ -72,14 +92,14 @@ The meaning of the setting:
 ## 3. Size, and what counts
 
 **The size** of a team is: the lead, every member (whatever its state — a finished member can be
-woken by a message, so it keeps its place), and every `task` helper running in the team, started by
+woken by a message, so it keeps its place), and every `task` subagent running in the team, started by
 any agent in it (the CLI's background `task` jobs and foreground ones alike, while they run).
 A stopped member does not count: stopping is how room is made.
 
 - **`team_add` is refused past `max`** — when the size would become more than `max`, with
   `team.texts.full`. So is `task`, while a team is offered: with `full`, or with `one_agent` when
   `max` is 1 and the lead has no member. A `job_send` that would start a finished background
-  `task` helper again is refused the same way.
+  `task` subagent again is refused the same way.
 - **Lowering `max` stops nobody.** Members already there stay; nothing new is added until the size
   is under the new `max`.
 - The size is across **the whole tree**: a member that leads adds within the same `max` as the lead.
@@ -90,8 +110,9 @@ A stopped member does not count: stopping is how room is made.
   machinery (a hidden chat on LLeMbas with its `chat_jobs` row, a subagent job in the CLI), running
   unattended — no `ask_user`, the chat's own permission ceiling, its mode no looser than its
   manager's (`permission/modes.json`).
-- **Its agent** is `general` (the default: it does the work) or `explore` (read-only); `model` is
+- **Its agent** is `general` (the default: it does the work) or `explore` (read-only). **Its model**
-  as `task`'s — a model the person may use, the default being the caller's.
+  is the session's (the lead's) unless `model` gives one the person named in this session — as
+  `task`'s, refused with `subagents.texts` otherwise (`subagents.md` §2, §3).
 - **Its tools** are those its agent gives a subagent, plus `team_send` and `team_read`; one added
   with `lead: true` also has `team_add` and `team_stop`. They are fixed when it is made (§10 rule 4:
   a subagent's narrowing). `report` (`offered: subagent`) sends its manager a message before its
@@ -108,7 +129,8 @@ A stopped member does not count: stopping is how room is made.
   `unavailable` (§2); `all_name` (the name `all`); `bad_name` (not `team.names.pattern`); `taken`
   (the name of an agent on the team, `lead` included); `no_goal`; `full`. Then the member is added:
   `added`, or `added_waiting` when it has to wait for a slot (§8). An unknown `agent` or `model`
-  answers as `task` does.
+  answers as `task` does; then `model`'s refusals, `not_named`, `same_group`, `one_request`
+  (`subagents.md`), in that order — before anything is added.
 - **Members are not jobs to the model.** `job_list` does not show them and the `job_*` tools do not
   take them; the team tools do. (An implementation may run them on its job machinery.)
 - **States**: `working` (a turn is running), `waiting` (it has work, and waits for a slot, §8),
@@ -171,7 +193,7 @@ class.
 **The tool list does not change** when an agent becomes a manager (`context.md` §10, rule 4): the
 same tools are listed and the refused ones answer the refusal. An agent is a manager while it has
 a member — a finished one included; one whose members were all stopped is not, unless it is the
-lead and `min ≥ 2` with the team offered. `task` stays: a one-off helper is delegation, and counts
+lead and `min ≥ 2` with the team offered. `task` stays: a one-off subagent is delegation, and counts
 toward the size (§3).
 
 ## 7. Delivery, and waking
@@ -180,7 +202,7 @@ A message or a report reaches its recipient **as a marked message** — the jobs
 (`loop.json` `jobs`): while the recipient's turn runs, as a user message at its next step boundary
 (which does not start a turn: `context.md` §8); otherwise it **wakes** the recipient — a finished
 member, or the idle lead — as a new turn of its own, marked `machine`, with its context intact, as
-`job_send` restarts a finished helper. A woken agent needs a slot (§8). Several waiting at one step
+`job_send` restarts a finished subagent. A woken agent needs a slot (§8). Several waiting at one step
 boundary go in one user message, in room order, separated by a blank line.
 
 Every one begins with its mark, on a line of its own, then the text:
@@ -199,14 +221,16 @@ A session that is not open keeps them until it is resumed, as a job's note.
 ## 8. Capacity: a member waits for a slot
 
 A model serves `parallel` requests at once (`models/schema.json`). **An agent holds a slot of its
-model while its turn runs** — the lead, a member, a `task` helper. An agent with work (added, woken,
+model while its turn runs** — the lead, a member, a `task` subagent. An agent with work (added, woken,
 given a message while finished, the lead given the person's message) starts when its model has a
 free slot and none of its `capacity.group`'s other models is busy; otherwise it is `waiting`.
-`team_add` is never refused for capacity (only for size): its answer says the member waits.
+`team_add` is never refused for a full model (only for size, and for the model rules of
+`subagents.md` §3: another model on the lead's one-model server, or a one-request session model):
+its answer says the member waits.
 
 Waiting agents start **in order**: the lead first (the person waits on it), then the others in the
 order they began waiting. A slot frees when a turn ends or an agent is stopped. `task` keeps its own
-rule (`loop.json` `capacity`: refused where its model is full), and its helper holds a slot.
+rule (`loop.json` `capacity`: refused where its model is full), and its subagent holds a slot.
 
 ## 9. Stopping
 
@@ -229,6 +253,10 @@ agent not under it (`not_yours`). The answer, `stopped`, names the stopped in tr
 
 ## 11. Over the link
 
+*3.4.0: the tree and the room are carried — `acp.md` capability `tree`; what follows is how a device
+without it is still shown.*
+
+
 An agent chat's team runs on the device, as its jobs do. Spec 3.2.0 does not carry the room over
 ACP yet: a device with the `jobs` capability reports each member as a subagent job (`job` events,
 `description` being `<name>: <role>`), so the client's job strip shows them, and the room is shown

harness/tools/task.json

@@ -12,7 +12,7 @@
       "subagent_run"
     ]
   },
-  "description": "Hand a self-contained piece of work to a subagent, which works on its own and returns only its final report — keeping the searching and reading out of your context.\n\n- Use it for broad research (\"where is X handled, and what calls it?\") or a separable job. For one file you already know, just read it.\n- `prompt` must stand on its own: the subagent sees nothing of this conversation. Say what to find or do, and what to report back.\n- `background: true` returns at once with a job id and you carry on; the report comes to you as a message when it ends. Meanwhile job_output shows its progress, job_send gives it a message, job_stop stops it. Without it, the call waits for the report.\n- Independent tasks can run in the same turn. Subagents that change files at the same time should each have `worktree: true`, so they work on separate branches.\n- Agents: {{agents}}",
+  "description": "Hand a self-contained piece of work to a subagent, which works on its own and returns only its final report — keeping the searching and reading out of your context.\n\n- Use it for broad research (\"where is X handled, and what calls it?\") or a separable job. For one file you already know, just read it.\n- `prompt` must stand on its own: the subagent sees nothing of this conversation. Say what to find or do, and what to report back.\n- `background: true` returns at once with a job id and you carry on; the report comes to you as a message when it ends. Meanwhile job_output shows its progress, job_send gives it a message, job_stop stops it. Without it, the call waits for the report.\n- Independent tasks can run in the same turn. Subagents that change files at the same time should each have `worktree: true`, so they work on separate branches.\n- It runs on your model. Another model only when the person named it in this conversation; nothing else chooses one.\n- Agents: {{agents}}",
   "parameters": {
     "type": "object",
     "properties": {
@@ -38,7 +38,7 @@
         "type": "boolean"
       },
       "model": {
-        "description": "Run it on another configured model (connection/model, from the list in your instructions) — a second opinion, or a model better at this kind of work. Default: the agent's own model, else yours.",
+        "description": "Run it on another model — only one the person named in this conversation (its id or name, as in the list in your instructions). Default: your model.",
         "type": "string"
       }
     },

harness/tools/team_add.json

@@ -14,7 +14,7 @@
     "cli": [],
     "llembas": []
   },
-  "description": "Add a member to your team and start it at once, in the background, on its part of the work. You lead: members do the work while you plan, coordinate and review — once you have a member, tools that change things are refused to you.\n\n- `name` is how everyone addresses it: lowercase letters, digits and hyphens, after what it owns (`parser`, `api-tests`), unique in the team.\n- `role` is one line: what it owns. `task` is its brief and must stand on its own — it sees nothing of this conversation: what to do, when it is done, and what to report.\n- `goal` is the team's goal in a line; your first member needs it, and every member is told it.\n- `lead: true` lets it add members of its own and lead an area; it then manages as you do.\n- The person sets how many agents the team may have; past that this is refused. When the model is already serving as many requests as it can at once, the member waits for a slot and starts when one frees.\n- Its final reply comes to you as its report. Meanwhile team_send messages it, team_read shows the team and the room, team_stop stops it.\n- Agents: {{agents}}",
+  "description": "Add a subagent and start it at once, in the background, on its part of the work. Only you, the main agent, add subagents. You lead: subagents do the work while you plan, coordinate and review — once you have one, tools that change things are refused to you.\n\n- `name` is how everyone addresses it: lowercase letters, digits and hyphens, after what it owns (`parser`, `api-tests`). `role` is its position, one line: what it owns. `task` is its brief and must stand on its own — it sees nothing of this conversation: what to do, when it is done, and what to report.\n- `goal` is the team's goal in a line; your first subagent needs it, and every subagent is told it.\n- `lead: true` makes it a sub-manager: it directs the subagents you place under it (`parent`), reviews their reports and reports to you; it does not add or stop subagents itself.\n- `mode` is its permission mode: its parent's by default, never looser. With a stricter mode you may give `request` options — kinds of action (read, write, execute) it may ask you for with team_request; you decide with team_decide.\n- The person sets how many agents the team may have, you included; past that this is refused. When the model is already serving as many requests as it can at once, the subagent waits for a slot and starts when one frees.\n- It reports with team_finish (or its final reply). Meanwhile team_send messages it, team_read shows the tree and the room, team_assign gives it new work, team_replace replaces it, team_stop stops it.\n- Agents: {{agents}}",
   "parameters": {
     "type": "object",
     "properties": {
@@ -26,7 +26,7 @@
       "role": {
         "type": "string",
         "minLength": 1,
-        "description": "One line: what it owns"
+        "description": "Its position, one line: what it owns"
       },
       "task": {
         "type": "string",
@@ -39,7 +39,7 @@
       },
       "lead": {
         "type": "boolean",
-        "description": "May add members of its own and lead an area (default: false)"
+        "description": "A sub-manager: directs the subagents placed under it (default: false)"
       },
       "agent": {
         "type": "string",
@@ -47,7 +47,33 @@
       },
       "model": {
         "type": "string",
-        "description": "Run it on another model you may use (from the list in your instructions). Default: yours."
+        "description": "Run it on another model — only one the person named in this conversation (its id or name, as in the list in your instructions). Default: the session's model."
+      },
+      "parent": {
+        "type": "string",
+        "description": "The sub-manager to place it under (default: you, lead)"
+      },
+      "mode": {
+        "type": "string",
+        "enum": [
+          "auto",
+          "edit",
+          "manual",
+          "plan"
+        ],
+        "description": "Its permission mode: its parent's or stricter (default: its parent's)"
+      },
+      "request": {
+        "type": "array",
+        "items": {
+          "type": "string",
+          "enum": [
+            "read",
+            "write",
+            "execute"
+          ]
+        },
+        "description": "With a stricter mode only: kinds of action it may ask you for (team_request)"
       }
     },
     "required": [
@@ -58,5 +84,6 @@
   },
   "variables": {
     "agents": "the agents a member may be: `name — description`, separated by \"; \" (as task's)"
-  }
+  },
+  "offered": "main"
 }

harness/tools/team_assign.json

@@ -0,0 +1,51 @@
+{
+  "name": "team_assign",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "offered": "main",
+  "description": "Give a subagent that has not ended a new task: it is told at its next step, or woken. Optionally a new position (`role`), another sub-manager to work under (`parent`, or lead), or a stricter `mode`.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "name": {
+        "type": "string",
+        "minLength": 1
+      },
+      "task": {
+        "type": "string",
+        "minLength": 1
+      },
+      "role": {
+        "type": "string"
+      },
+      "parent": {
+        "type": "string"
+      },
+      "mode": {
+        "type": "string",
+        "enum": [
+          "auto",
+          "edit",
+          "manual",
+          "plan"
+        ]
+      }
+    },
+    "required": [
+      "name",
+      "task"
+    ]
+  }
+}

harness/tools/team_decide.json

@@ -0,0 +1,38 @@
+{
+  "name": "team_decide",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "offered": "main",
+  "description": "Decide a subagent's request (team_request): allow it or refuse it, with a note it is told. You may allow only what your own mode would allow you; anything beyond that goes to the person when the subagent tries.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "id": {
+        "type": "string",
+        "minLength": 1
+      },
+      "allow": {
+        "type": "boolean"
+      },
+      "note": {
+        "type": "string"
+      }
+    },
+    "required": [
+      "id",
+      "allow"
+    ]
+  }
+}

harness/tools/team_findings.json

@@ -0,0 +1,41 @@
+{
+  "name": "team_findings",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "description": "Findings kept beyond your context: `write` adds one (a fact, a result, where something is) to yours; `read` shows yours, or with `of` those of a subagent under you (the main agent: anyone's), ended ones included. Write as you go: when your context fills you have one turn left, and your findings are what the next agent starts from.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "action": {
+        "type": "string",
+        "enum": [
+          "write",
+          "read"
+        ]
+      },
+      "text": {
+        "type": "string",
+        "description": "write: the finding"
+      },
+      "of": {
+        "type": "string",
+        "description": "read: a name or an id (default: yours)"
+      }
+    },
+    "required": [
+      "action"
+    ]
+  }
+}

harness/tools/team_finish.json

@@ -0,0 +1,31 @@
+{
+  "name": "team_finish",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "offered": "subagent",
+  "description": "Your task is done: send your report to your parent and end. Make the report complete on its own — the result, then the evidence (file:line), then anything you could not settle. Save anything worth keeping beyond the report with team_findings first. Nothing runs after this call.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "report": {
+        "type": "string",
+        "minLength": 1
+      }
+    },
+    "required": [
+      "report"
+    ]
+  }
+}

harness/tools/team_read.json

@@ -14,7 +14,7 @@
     "cli": [],
     "llembas": []
   },
-  "description": "The team and its room: every agent (name, role, who leads it, state, model, last activity), and the room's messages after `since` that are yours to see — those you sent or received and those of the members under you; the lead sees them all. Messages are numbered: pass the last number you saw as `since` to read only what is new.",
+  "description": "The tree and its room: every agent (name, id, position, parent, state, model, mode, context use, last activity), open slots, subagents that have ended, and the room's messages after `since` that are yours to see — those you sent or received and those of the subagents under you; the main agent sees them all. Messages are numbered: pass the last number you saw as `since` to read only what is new.",
   "parameters": {
     "type": "object",
     "properties": {

harness/tools/team_replace.json

@@ -0,0 +1,34 @@
+{
+  "name": "team_replace",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "task",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "offered": "main",
+  "description": "Replace a subagent — one that ran out of context, or one that is stuck: it ends (stopped, if it has not already), and a new one with the same name, position, parent, mode and model takes its place, briefed with `task` (default: the old task) and the old one's findings. The subagents under it move under the new one.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "name": {
+        "type": "string",
+        "minLength": 1
+      },
+      "task": {
+        "type": "string"
+      }
+    },
+    "required": [
+      "name"
+    ]
+  }
+}

harness/tools/team_request.json

@@ -0,0 +1,41 @@
+{
+  "name": "team_request",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "interact",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "offered": "subagent",
+  "description": "Ask the main agent to allow a call your mode does not, of a kind your request options name. Say the tool, the permission pattern you need (a command such as `git push *`, or a path) and why. Its decision comes to you as a message; once allowed, make the call again.",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "tool": {
+        "type": "string",
+        "minLength": 1
+      },
+      "pattern": {
+        "type": "string",
+        "minLength": 1
+      },
+      "why": {
+        "type": "string",
+        "minLength": 1
+      }
+    },
+    "required": [
+      "tool",
+      "pattern",
+      "why"
+    ]
+  }
+}

harness/tools/team_send.json

@@ -14,7 +14,7 @@
     "cli": [],
     "llembas": []
   },
-  "description": "Send a message through your team's room: to one or more agents by name, or to `all`. It reaches each at its next step, or wakes one that has finished, its context intact. Message the agent who owns a thing; use `all` only for what everyone needs. The team's lead is `lead`; your manager is reached by its name.",
+  "description": "Send a message through the team's room: to one or more agents by name, to `parent` (the agent you report to), or to `all`. It reaches each at its next step, or wakes one that has finished, its context intact. Message the agent who owns a thing; use `all` only for what everyone needs. The main agent is `lead`.",
   "parameters": {
     "type": "object",
     "properties": {
@@ -24,7 +24,7 @@
           "type": "string"
         },
         "minItems": 1,
-        "description": "Names, or [\"all\"]"
+        "description": "Names, [\"parent\"], or [\"all\"]"
       },
       "message": {
         "type": "string",

harness/tools/team_stop.json

@@ -14,18 +14,19 @@
     "cli": [],
     "llembas": []
   },
-  "description": "Stop a member of your team, and the members under it, where they are: their work stops and they leave the team, which makes room for others. Only members under you, directly or through your members; `all` stops every one of them. Nothing they had not reported comes back — ask for a report with team_send first when you need it.",
+  "description": "Stop a subagent, and the subagents under it, where they are: their work stops and they end, which makes room for others. Only you, the main agent, stop subagents; `all` stops every one. Nothing they had not reported comes back — ask for a report with team_send first, or read their findings with team_findings.",
   "parameters": {
     "type": "object",
     "properties": {
       "name": {
         "type": "string",
         "minLength": 1,
-        "description": "A member under you, or all"
+        "description": "A subagent's name, or all"
       }
     },
     "required": [
       "name"
     ]
-  }
+  },
+  "offered": "main"
 }

harness/tools/team_whoami.json

@@ -0,0 +1,22 @@
+{
+  "name": "team_whoami",
+  "scope": "shared",
+  "sessions": "both",
+  "risk": "read",
+  "exclusive": false,
+  "block": "tool",
+  "purpose": false,
+  "requires": [
+    "team"
+  ],
+  "unavailable": "The person allows one agent in this chat: there is no team. Do the work yourself.",
+  "aliases": {
+    "cli": [],
+    "llembas": []
+  },
+  "description": "Who you are on the team: your id, name, position, parent, task, permission mode and request options, model, limits, and how full your context is.",
+  "parameters": {
+    "type": "object",
+    "properties": {}
+  }
+}

install.sh

@@ -324,5 +324,5 @@ say ""
 say "Done. Open a new shell (or: source your rc file), then run: lembas"
 [ -n "$extra" ] && say "Aliases: $(printf '%s' "$extra" | sed -n "s/^alias \([a-z]*\)=.*/\1/p" | paste -sd' ')"
 if [ "$starter" -eq 1 ] && [ -z "$(sed -n 's/^model: *//p' "$CONFIG/config.yaml")" ]; then
-  say "Next: add a connection to $CONFIG/connections.yaml and set model: in config.yaml."
+  say "Next: lembas login <your LLeMbas server address>"
 fi

package.json

@@ -1,6 +1,6 @@
 {
   "name": "lembas-cli",
-  "version": "2.0.0-beta.3",
+  "version": "2.0.0-beta.6",
   "description": "LLeMbas CLI: a terminal coding agent and project manager for a LLeMbas server — OpenTUI, memory, skills, MCP, git-native.",
   "license": "MIT",
   "private": true,

schema/config.schema.json

@@ -894,17 +894,17 @@
       ]
     },
     "agents": {
-      "description": "Teams (harness team.md): how many agents a session's team may have. Offered only where subagents are on and the model serves two requests at once. /agents changes it for the session.",
+      "description": "Subagents (harness team.md, subagents.md): how many agents a session may have at work, counting the session's own agent — Subagents: Min 1, Max 4 by default; 1 is this agent alone. Offered where subagents are on and a model could serve one: the session's, when it serves two requests at once, or another model off its one-model server. /team size and /effort (one range, Min and Max) change it for the session.",
       "type": "object",
       "properties": {
         "min": {
-          "description": "The fewest agents a team works with, the lead included (default 1). 2 or more: the lead works through a team from the start.",
+          "description": "Subagents, Min: the agents at work at least, this agent included (default 1 — 1 is this agent alone). 2 or more: it works through subagents from the start.",
           "type": "integer",
           "minimum": 1,
           "maximum": 9007199254740991
         },
         "max": {
-          "description": "The most agents a team may have, the lead included (default 4). 1: one agent, no team and no helper.",
+          "description": "Subagents, Max: the agents at work at most, this agent included (default 4 — this agent and up to 3 subagents). 1: this agent alone, no subagents.",
           "type": "integer",
           "minimum": 1,
           "maximum": 9007199254740991

schema/connections.schema.json

@@ -210,11 +210,11 @@
                   "maxLength": 300
                 },
                 "single_session": {
-                  "description": "The server serves one request at a time: a subagent on this same model would take its only slot and push the session's cached prompt out, so it is refused. The same as parallel: 1.",
+                  "description": "The server serves one request at a time: a subagent on this same model would take its only slot and push the session's cached prompt out, so it is refused (name a model on another server instead). The same as parallel: 1.",
                   "type": "boolean"
                 },
                 "parallel": {
-                  "description": "How many requests the model's server serves at once (capacity.parallel); wins over single_session. Absent with no single_session: no limit (a hosted API). A team is offered only on a model that serves at least 2.",
+                  "description": "How many requests the model's server serves at once (capacity.parallel); wins over single_session. Absent with no single_session: no limit (a hosted API). The session's model is its own subagents' model only when it serves at least 2.",
                   "type": "integer",
                   "exclusiveMinimum": 0,
                   "maximum": 9007199254740991

src/acp/agent.ts

@@ -42,7 +42,7 @@ import { CWD_META, REMOTE_META, Store } from "../session/store.ts"
 import type { Message } from "../provider/types.ts"
 import type { Hub } from "./hub.ts"
 import { INVALID_PARAMS, METHOD_NOT_FOUND, RpcError, type Peer } from "./rpc.ts"
-import { jobList, jobOutput, jobSend, jobStop, jobTranscript, storedJobs } from "./jobs.ts"
+import { jobList, jobOutput, jobSend, jobStop, jobTranscript, storedJobs, teamFindings, teamStop, teamTranscript, teamView, treeOf } from "./jobs.ts"
 import { isNote } from "../session/prefix.ts"
 
 export const ACP_VERSION = 1
@@ -273,6 +273,12 @@ export const CAPABILITIES = [
   // Harness spec 3.0.0: the show tool, offered in a session the client opened when the client
   // said it can show files; the device sends the file with `_lembas/show`.
   "show",
+  // Harness spec 3.4.0: the progress phases (the `phase` event, `progress` in status); the subagent
+  // tree (the `team` event with the tree, `_lembas/team*`, `subagent` on an approval); the chat's
+  // Subagents setting read from `_meta.lembas.agents` (an agent chat's subagents are this device's).
+  "progress",
+  "tree",
+  "agents",
 ] as const
 export const STOP: Record<string, string> = { stop: "end_turn", steps: "max_turn_requests", budget: "max_tokens", cancelled: "cancelled", error: "end_turn" }
 
@@ -306,6 +312,8 @@ export function permissionParams(sessionId: string, req: AskRequest, o: { patien
         ...(o.patient ? { patient: true } : {}),
         // Which subagent asks (capability jobs), so the card can say — it may ask while the session itself is idle.
         ...(req.source ? { job: { id: req.source.jobId, agent: req.source.agent, description: req.source.description } } : {}),
+        // Spec 3.4.0 (capability tree): a subagent of the tree whose call only the person can allow.
+        ...(req.source?.subagent ? { subagent: req.source.subagent } : {}),
       },
     },
   }
@@ -468,6 +476,10 @@ export class AcpAgent {
     peer.handle("_lembas/job/stop", (p) => this.job("stop", p))
     peer.handle("_lembas/job/send", (p) => this.job("send", p))
     peer.handle("_lembas/job/transcript", (p) => this.job("transcript", p))
+    peer.handle("_lembas/team", (p) => this.tree("team", p))
+    peer.handle("_lembas/team/transcript", (p) => this.tree("team_transcript", p))
+    peer.handle("_lembas/team/findings", (p) => this.tree("team_findings", p))
+    peer.handle("_lembas/team/stop", (p) => this.tree("team_stop", p))
     peer.on("_lembas/terminal/input", (p) => this.terminalInput(p))
     peer.on("_lembas/terminal/resize", (p) => this.terminalResize(p))
     peer.on("_lembas/terminal/close", (p) => this.terminalClose(String(p?.terminalId ?? "")))
@@ -675,6 +687,8 @@ export class AcpAgent {
     const effort = effortOf(p?.effort)
     // A level the model does not offer is read as its start level; the answer says the one used.
     if (effort !== undefined) s.app.engine.effort = readLevel(s.app.engine.model, effort)
+    // Spec 3.4.0: the chat's Subagents setting, changed between turns.
+    if (p?.agents) this.takeAgents(s, { agents: p.agents })
     return { model: s.app.engine.model.ref, effort: s.app.engine.effort }
   }
 
@@ -724,6 +738,19 @@ export class AcpAgent {
   private takeContext(s: Session, p: any) {
     const ctx = contextOf(p?._meta)
     if (ctx) s.app.setContext(ctx, this.instanceConnection)
+    this.takeAgents(s, p?._meta?.lembas)
+  }
+
+  /** Spec 3.4.0 (capability agents): the chat's Subagents setting, sent with every turn by the
+   *  instance — absent (an older one, an editor on serve --stdio): the device's own, as before.
+   *  `said`: the person's earlier words in the chat. */
+  private takeAgents(s: Session, m: any) {
+    if (!this.lembas) return
+    const team = s.app.team.team
+    const a = m?.agents
+    if (Number.isInteger(a?.min) && Number.isInteger(a?.max) && a.min >= 1 && a.max >= 1) team.setSetting(Math.min(a.min, a.max), a.max)
+    if (m?.subagents === false) team.subagents = false
+    if (Array.isArray(m?.said) && !s.app.engine.saidBefore.length) s.app.engine.saidBefore = m.said.filter((x: unknown): x is string => typeof x === "string")
   }
 
   /** The session's plan, as a `plan` update — after session/new and after opening from the
@@ -1067,6 +1094,8 @@ export class AcpAgent {
         ...(s.turns.last ? { lastTurnId: s.turns.last } : {}),
         plan: s.app.engine.plan ?? null,
         jobs: jobList(s.app),
+        progress: s.app.engine.progress.current,
+        tree: treeOf(s.app),
       }
     const row = (this.store ??= new Store()).session(id)
     // Not open: the last turn it finished, as kept across a restart, its plan and its jobs as stored.
@@ -1091,6 +1120,24 @@ export class AcpAgent {
     return { jobs: jobList(this.get(p).app) }
   }
 
+  /** `_lembas/team`, `/transcript`, `/findings`, `/stop` (spec 3.4.0, capability tree): the
+   *  session's subagent tree — the terminal's through the hub, as the jobs go. */
+  private async tree(op: "team" | "team_transcript" | "team_findings" | "team_stop", p: any) {
+    const t = this.terminalFor(p)
+    if (t) return t.client.request("hub/job", { sessionId: t.sessionId, op, id: p?.id, since: p?.since }, 15_000)
+    const app = this.get(p).app
+    switch (op) {
+      case "team":
+        return teamView(app, p?.since)
+      case "team_transcript":
+        return teamTranscript(app, p?.id)
+      case "team_findings":
+        return teamFindings(app, p?.id)
+      case "team_stop":
+        return teamStop(app, p?.id)
+    }
+  }
+
   /** `_lembas/job/output`, `/stop`, `/send`, `/transcript` (acp/jobs.ts): one job of the session. */
   private async job(op: "output" | "stop" | "send" | "transcript", p: any) {
     const t = this.terminalFor(p)
@@ -1291,7 +1338,7 @@ export class AcpAgent {
   private forward(s: Session, e: Event) {
     // A model change says whether the model is this link's instance's: its ref is the
     // provider's now (`deepseek/x`), so the ref alone no longer tells.
-    const event = e.type === "model" ? { ...e, instance: this.isInstance(e.connection) } : e
+    const event = e.type === "model" ? { ...e, instance: this.isInstance(e.connection) } : e.type === "team" ? { ...e, tree: treeOf(s.app) } : e
     if (this.lembas) this.peer.notify("_lembas/event", { sessionId: s.id, event })
     switch (e.type) {
       case "text":

src/acp/jobs.ts

@@ -5,6 +5,7 @@
 import type { App } from "../app.ts"
 import { jobRegistry, type Job, type JobInfo } from "../jobs/registry.ts"
 import type { Store } from "../session/store.ts"
+import type { TreeInfo } from "../team/team.ts"
 import { historyTurns } from "./agent.ts"
 import { INVALID_PARAMS, RpcError } from "./rpc.ts"
 
@@ -79,3 +80,48 @@ export function storedJobs(store: Store, sessionId: string): JobInfo[] {
   }
   return [...list.filter((j) => j.state === "running"), ...list.filter((j) => j.state !== "running")]
 }
+
+// ── Spec 3.4.0 (acp.md capability `tree`): the subagent tree of a session ────────────────────
+
+/** The tree (subagents.md §12), or null for a session with no subagents. */
+export function treeOf(app: App): TreeInfo | null {
+  const t = app.team.team
+  if (!t.offered && !t.hasMembers() && !t.ended.length) return null
+  return t.info()
+}
+
+/** `_lembas/team`: the tree and the whole room above `since`. */
+export function teamView(app: App, since: unknown) {
+  const n = typeof since === "number" && since >= 0 ? since : 0
+  const t = app.team.team
+  return { tree: treeOf(app), room: t.room.filter((m) => m.n > n).map((m) => ({ n: m.n, from: m.from, to: m.to, recipients: m.recipients, text: m.text, time: m.time, report: m.report, kind: m.kind })) }
+}
+
+function subagentOf(app: App, id: unknown) {
+  const key = String(id ?? "")
+  const a = app.team.team.find(key)
+  if (!a || a.id !== key || a.id === "lead") throw new RpcError(INVALID_PARAMS, `this session has no subagent ${key}`)
+  return a
+}
+
+/** `_lembas/team/transcript`: a subagent's session, read-only. */
+export function teamTranscript(app: App, id: unknown) {
+  const a = subagentOf(app, id)
+  const job = a.jobId ? app.registry.get(app.jobsSessionId(), a.jobId) : undefined
+  const messages = job?.subagent?.engine.messages ?? (job?.childSessionId && app.store ? app.store.messages(job.childSessionId) : [])
+  return { id: a.id, state: a.state, turns: historyTurns(messages, app.engine.o.toolCtx.cwd) }
+}
+
+/** `_lembas/team/findings`. */
+export function teamFindings(app: App, id: unknown) {
+  const a = subagentOf(app, id)
+  return { id: a.id, findings: a.findings.map((f) => ({ ...f })) }
+}
+
+/** `_lembas/team/stop`: the person stops a live subagent and its subtree. */
+export function teamStop(app: App, id: unknown) {
+  const a = subagentOf(app, id)
+  if (a.endState) throw new RpcError(INVALID_PARAMS, `${a.id} has already ended`)
+  const gone = app.team.personStop(a.name)
+  return { stopped: app.team.team.ended.slice(-gone.length).map((x) => x.id) }
+}

src/acp/share.ts

@@ -16,7 +16,7 @@ import type { PlanReply, PlanRequest } from "../tool/plan_submit.ts"
 import type { QuestionReply, QuestionRequest } from "../tool/ask_user.ts"
 import { readLevel } from "../provider/effort.ts"
 import { clampMode, effortOf, markRemote, permissionParams, planParams, planReplyFrom, questionParams, questionReplyFrom, replyFrom, STOP, type PromptResult } from "./agent.ts"
-import { jobList, jobOutput, jobSend, jobStop, jobTranscript } from "./jobs.ts"
+import { jobList, jobOutput, jobSend, jobStop, jobTranscript, teamFindings, teamStop, teamTranscript, teamView, treeOf } from "./jobs.ts"
 import type { SessionContext } from "../app.ts"
 import { sessionTurnFile, sessionTurns, type TurnLog } from "./turns.ts"
 import { dropAttachments, preparePrompt } from "./prompt.ts"
@@ -454,7 +454,7 @@ export class Share {
       return
     }
     if (this.sharedId && this.hub && this.app.engine.sessionId === this.sharedId) {
-      this.hub.notify("hub/event", { sessionId: this.sharedId, event: e })
+      this.hub.notify("hub/event", { sessionId: this.sharedId, event: e.type === "team" ? { ...e, tree: treeOf(this.app) } : e })
       if (e.type === "task" && e.state === "end") this.sendCommands()
     }
   }
@@ -520,6 +520,15 @@ export class Share {
           return jobSend(this.app, p?.id, p?.text)
         case "transcript":
           return jobTranscript(this.app, p?.id)
+        // Spec 3.4.0 (capability tree): the subagent tree, through the hub as the jobs go.
+        case "team":
+          return teamView(this.app, p?.since)
+        case "team_transcript":
+          return teamTranscript(this.app, p?.id)
+        case "team_findings":
+          return teamFindings(this.app, p?.id)
+        case "team_stop":
+          return teamStop(this.app, p?.id)
         default:
           throw new RpcError(INVALID_PARAMS, `no such job operation: ${String(p?.op ?? "")}`)
       }

src/app.ts

@@ -16,6 +16,7 @@ import type { Mode } from "./config/schema.ts"
 import { DEFAULT_RULES, toRules } from "./permission/evaluate.ts"
 import { hardlineRules } from "./permission/hardline.ts"
 import { clientFor, familyOf, resolveModel } from "./provider/index.ts"
+import { subagentModel, type ModelInfo } from "./team/subagents.ts"
 import { assembleSystem, instructionFiles, knowledgeScope, modeText } from "./prompt/assemble.ts"
 import type { Changing } from "./session/prefix.ts"
 import { describeRepo, repoState } from "./git/repo.ts"
@@ -45,7 +46,8 @@ import { LIBRARY_SERVER, mcpInstructions } from "./mcp/index.ts"
 import type { McpServer } from "./config/schema.ts"
 import { embedderFor } from "./library/embed.ts"
 import { isLevel, levelsFor, readLevel, type Level } from "./provider/effort.ts"
-import { readingBashByDefault, Team, TEAM, TEAM_TOOLS, type Capacity, type TeamAgent } from "./team/team.ts"
+import { memberTools as memberToolsOf, readingBashByDefault, Team, TEAM, TEAM_TOOLS, type Capacity, type TeamAgent } from "./team/team.ts"
+import { whoDecides } from "./team/delegation.ts"
 import { TeamRuntime, type MemberHandle } from "./team/runtime.ts"
 
 export interface AppOptions {
@@ -202,8 +204,15 @@ function openInBrowser(url: string): boolean {
   }
 }
 
+/** A model as the subagent rules see it (subagents.md): its names and what its server can do at once. */
+function modelInfo(m: ResolvedModel): ModelInfo {
+  return { ref: m.ref, id: m.id, name: m.spec.name, capacity: { parallel: m.spec.parallel, single_session: m.spec.single_session, group: m.spec.group } }
+}
+
 /** Tools the spec offers to subagents only (`offered: subagent`): never a session's own model's. */
 export const SUBAGENT_ONLY = new Set(Object.values(TOOL_SPECS).filter((t) => t.offered === "subagent").map((t) => t.name))
+/** Spec 3.4.0: tools the spec offers to the main agent only (`offered: main`): never a subagent's. */
+export const MAIN_ONLY_TOOLS = new Set(Object.values(TOOL_SPECS).filter((t) => t.offered === "main").map((t) => t.name))
 
 const MODE_RANK: Record<Mode, number> = { auto: 0, edit: 1, manual: 2, plan: 3 }
 export function stricterMode(a: Mode | undefined, b: Mode): Mode {
@@ -364,7 +373,7 @@ export function createApp(o: AppOptions): App {
     if (src.instance) named[`Instructions from ${src.instance.name}`] = src.instance.instructions
     return { memory: src.memory ?? "", personality: src.personality ?? "", files: named, mcp: src.mcp, source: src }
   }
-  const system = (m: Mode, current: ResolvedModel, ch: Changing, subagent?: { name: string; instructions: string; report: boolean; skills: boolean; recall: boolean }, at?: { cwd: string; root: string }, plan?: Plan, member?: { name: string; manager: string; role: string; goal: string; lead: boolean }) => {
+  const system = (m: Mode, current: ResolvedModel, ch: Changing, subagent?: { name: string; instructions: string; report: boolean; skills: boolean; recall: boolean }, at?: { cwd: string; root: string }, plan?: Plan, member?: { id: string; name: string; manager: string; role: string; goal: string; lead: boolean }) => {
     const src = ch.source as Source
     return assembleSystem({
       modelRef: current.ref,
@@ -473,13 +482,22 @@ export function createApp(o: AppOptions): App {
     // its server can take it beside the session's and beside every subagent already running
     // (src/session/capacity.ts), before anything is made.
     const member = opts.member
-    const m = opts.model ? resolveModel(loaded, opts.model) : spec.model ? resolveModel(loaded, spec.model) : engine.model
+    // Spec 3.3.0 (subagents.md): the session's model, unless the person named another; an agent
+    // file's own `model` is not used. A member's model was judged by team_add.
+    let m: ResolvedModel
+    if (member) m = opts.model ? resolveModel(loaded, opts.model) : engine.model
+    else {
+      const wanted = opts.model ? resolveModel(loaded, opts.model) : undefined
+      const r = subagentModel(modelInfo(engine.model), wanted && modelInfo(wanted), engine.personSaid())
+      if ("refused" in r) throw new ToolError(r.refused)
+      m = wanted && wanted.ref === r.model ? wanted : engine.model
+    }
     // A team's member is not refused for capacity: the team gave it its slot (team.md §8).
     if (!member) {
       const running = registry.running(jobsSessionId()).filter((j) => j.kind === "subagent" && j.subagent && !j.team)
       const onIt = running.filter((j) => j.subagent!.engine.model.ref === m.ref).length
       const refused = capacityRefusal(engine.model, m, onIt) || running.map((j) => capacityBeside(j.subagent!.engine.model, m, onIt)).find(Boolean)
-      if (refused) throw new ToolError(`${refused} Do this part yourself, or hand it to a model on another connection.`)
+      if (refused) throw new ToolError(`${refused} Wait for one to finish, or do this part yourself.`)
     }
     // A checkout and branch of its own, when asked for (the task's worktree, the agent's isolation).
     let wt: Worktree | undefined
@@ -508,13 +526,17 @@ export function createApp(o: AppOptions): App {
         }
       : perm
     // `report` is every subagent's: its way of speaking to the main agent before its final reply.
-    // A team member also has team_send and team_read, and one that may lead team_add and team_stop
+    // A subagent of the tree also has its team tools (spec 3.4.0 §8: whoami, read, send, findings,
-    // (team.md §4); a task helper has none of them.
+    // finish, and request with request options) — never the main agent's (`offered: main`); a task
-    const memberTools = member ? (member.leads ? [...TEAM_TOOLS] : ["team_read", "team_send"]) : []
+    // subagent has none of them.
+    const memberTools = member ? memberToolsOf(member) : []
+    const isTeamTool = (n: string) => (TEAM_TOOLS as readonly string[]).includes(n)
     const childTools = [
-      ...tools.filter((t) => !NOT_FOR_SUBAGENTS.has(t.name) && (!(TEAM_TOOLS as readonly string[]).includes(t.name) || memberTools.includes(t.name)) && (!spec.tools || spec.tools.includes(t.name) || memberTools.includes(t.name))),
+      ...tools.filter((t) => !NOT_FOR_SUBAGENTS.has(t.name) && !MAIN_ONLY_TOOLS.has(t.name) && (!isTeamTool(t.name) || memberTools.includes(t.name)) && (!spec.tools || spec.tools.includes(t.name) || memberTools.includes(t.name))),
-      ...BUILTIN_TOOLS.filter((t) => SUBAGENT_ONLY.has(t.name)),
+      ...BUILTIN_TOOLS.filter((t) => SUBAGENT_ONLY.has(t.name) && (!isTeamTool(t.name) || memberTools.includes(t.name))),
     ]
+    // Its mode: the one the main agent gave it, made stricter by its agent's own (explore reads).
+    if (member) member.mode = stricterMode(spec.mode, member.mode)
     const description = member ? `${member.name}: ${member.role}` : (opts.description ?? prompt.split("\n")[0]!.slice(0, 60))
     const ownerSession = jobsSessionId()
     const childRow = store?.createSession(at?.root ?? project.root, m.ref, engine.sessionId)
@@ -524,7 +546,7 @@ export function createApp(o: AppOptions): App {
     let job: Job
     // The asker is the session's; a child's ask says which subagent it comes from.
     const askerFor = (): Asker => ({
-      ask: (req) => o.asker.ask({ ...req, source: { jobId: job.id, agent: spec.name, description } }),
+      ask: (req) => o.asker.ask({ ...req, source: { jobId: job.id, agent: spec.name, description, ...(member ? { subagent: { id: member.id, name: member.name, position: member.role } } : {}) } }),
       get question() {
         return o.asker.question
       },
@@ -540,9 +562,9 @@ export function createApp(o: AppOptions): App {
       tools: childTools,
       store,
       sessionId: childRow?.id,
-      // An agent's own mode can make it stricter than its parent, never looser.
+      // An agent's own mode can make it stricter than its parent, never looser. A subagent of the
-      // A team member's is no looser than its manager's (team.md §4): a sub-manager's own mode.
+      // tree runs in the mode the main agent gave it (spec 3.4.0 §7), set above.
-      mode: stricterMode(spec.mode, (member?.manager && member.manager !== TEAM.names.lead ? teamRt.handle(member.manager)?.engine.mode : undefined) ?? engine.mode),
+      mode: member ? member.mode : stricterMode(spec.mode, engine.mode),
       effort: m.ref === engine.model.ref ? engine.effort : startEffort(m),
       maxSteps: loaded.config.limits?.subagent_steps ?? LOOP.jobs.subagent_steps.default,
       compaction,
@@ -559,6 +581,13 @@ export function createApp(o: AppOptions): App {
               fix: () => teamRt.team.offered,
               offered: () => teamRt.team.offered,
               refuse: (tool: Tool, args: unknown) => teamRt.refuse(member.name, tool, args),
+              // Spec 3.4.0 §7: itself, the main agent through a request, or the person.
+              permit: (request, ctx) => {
+                const a = teamRt.team.get(member.name)
+                return whoDecides(request, { ctx, mode: a?.mode ?? member.mode, main: engine.mode, request: a?.request ?? [], grants: a?.grants ?? [], unattended: o.unattended })
+              },
+              contextFull: (used: number, window: number) => (window ? teamRt.contextFull(member.name) : "go"),
+              ended: () => teamRt.hasEnded(member.name),
             },
           }
         : {}),
@@ -581,7 +610,7 @@ export function createApp(o: AppOptions): App {
           },
           at,
           plan,
-          member ? { name: member.name, manager: member.manager ?? TEAM.names.lead, role: member.role, goal: member.goal, lead: member.leads } : undefined,
+          member ? { id: member.id, name: member.name, manager: member.manager ?? TEAM.names.lead, role: member.role, goal: member.goal, lead: member.leads } : undefined,
         ),
     })
     // Its own background commands wake it at its next step while it runs; once it has ended they
@@ -597,14 +626,14 @@ export function createApp(o: AppOptions): App {
       const end = registry.finishSubagent(job, { reason, report: text, ...(error ? { error } : {}), ...(worktree ? { worktree } : {}) })
       return { text: worktree ? `${text}\n\n[worktree] ${worktree}` : text, steps: job.steps ?? 0, tools: job.tools ?? 0, reason, ...(error ? { error } : {}), ...(end.stopped ? { stopped: end.stopped } : {}) }
     }
-    // A task helper started inside a team counts toward its size and holds a slot while it runs
+    // A task subagent started inside a team counts toward its size and holds a slot while it runs
     // (team.md §3, §8).
     const counted = (p: Promise<SubagentResult>) => {
       if (member || !teamRt.team.offered) return p
-      const id = teamRt.team.helperStart(TEAM.names.lead, m.ref)
+      const id = teamRt.team.taskStart(TEAM.names.lead, m.ref)
-      return p.finally(() => teamRt.team.helperDone(id))
+      return p.finally(() => teamRt.team.taskDone(id))
     }
-    const run = (text: string, machine = false) => counted(child.prompt(text, [], text, machine ? { machine: true } : {}).then(finish, (e) => (job.lastError = (e as Error).message, finish("error"))))
+    const run = (text: string, machine = false, steps?: number) => counted(child.prompt(text, [], text, { ...(machine ? { machine: true as const } : {}), ...(steps ? { maxSteps: steps } : {}) }).then(finish, (e) => (job.lastError = (e as Error).message, finish("error"))))
     const handle: SubagentHandle = {
       bus: childBus,
       engine: child,
@@ -616,15 +645,26 @@ export function createApp(o: AppOptions): App {
     if (member) {
       // A member's turn that ends is its report (team.md §5): the team, not a job note, carries it.
       job.team = member.name
-      const memberRun = (text: string, machine: boolean) => {
+      member.jobId = job.id
+      // Its progress and context use, for the tree (spec 3.4.0 §12).
+      childBus.on((e) => {
+        if (e.type === "phase") {
+          teamRt.team.setProgress(member.name, e.progress)
+          teamRt.notify()
+        } else if (e.type === "usage" && !e.child && e.context) {
+          teamRt.team.setContext(member.name, e.used ?? e.usage.input + e.usage.output, e.context)
+          teamRt.notify()
+        }
+      })
+      const memberRun = (text: string, machine: boolean, steps?: number) => {
         if (job.state !== "running") registry.restartSubagent(job)
-        void run(text, machine).then((r) => {
+        void run(text, machine, steps).then((r) => {
           const leftover = [...child.takeNotes(), ...child.takeInbox()]
           const reason = r.reason === "error" ? `error — ${r.error ?? "unknown"}` : r.reason
           if (r.reason !== "cancelled") teamRt.memberDone(member.name, r.text, reason, leftover)
         })
       }
-      const mh: MemberHandle = { engine: child, run: (text) => memberRun(text, true), cancel: () => child.cancel() }
+      const mh: MemberHandle = { engine: child, run: (text) => memberRun(text, true), cancel: () => child.cancel(), final: (text) => memberRun(text, true, TEAM.context.final_steps) }
       teamRt.attach(member.name, mh)
       memberRun(prompt, false)
       return { id: job.id, done: Promise.resolve({ text: "", steps: 0, tools: 0, reason: "stop" }), member: mh }
@@ -683,6 +723,8 @@ export function createApp(o: AppOptions): App {
       return {}
     }
   }
+  // Every model the person may use: the server's, as the login read them (subagents.md §4).
+  const usableModels = () => loaded.refs.map((r) => r.ref)
   // The size as set while the process runs (any scope): a new or resumed session keeps it.
   const teamSize: { min?: number; max?: number } = {}
   const newTeam = () =>
@@ -693,8 +735,11 @@ export function createApp(o: AppOptions): App {
       subagents: loaded.config.subagents !== false,
       model: engine.model.ref,
       capacity: capacityOf,
+      models: usableModels,
       leadState: "idle",
       readingBash: (command) => readingBashByDefault(command, project.root),
+      mode: engine.mode,
+      steps: { main: loaded.config.limits?.steps ?? LOOP.steps.default, sub: loaded.config.limits?.subagent_steps ?? LOOP.jobs.subagent_steps.default },
     })
   const teamRt: TeamRuntime = new TeamRuntime({
     team: newTeam(),
@@ -707,11 +752,19 @@ export function createApp(o: AppOptions): App {
         teamRt.memberDone(member.name, "", `error — ${(e as Error).message}`, [])
       })
     },
-    lead: { note: (text) => engine.note(text), wake: (text) => wake(text, shownLine(text)) },
+    lead: {
+      note: (text) => engine.note(text),
+      wake: (text) => wake(text, shownLine(text)),
+      waiting: (model) => engine.progress.set({ phase: "waiting_for_model", model: engine.model.spec.name ?? model, reason: "slots" }),
+    },
+    tools: () => tools,
     resolve: (agent, model) => {
       if (!agentList.some((a) => a.name === agent)) return { error: `There is no agent "${agent}". The agents are: ${agentList.map((a) => a.name).join(", ")}.` }
       try {
-        return { model: model ? resolveModel(loaded, model).ref : engine.model.ref }
+          // Spec 3.3.0 (subagents.md): the session's model, unless the person named another.
+        const wanted = model ? resolveModel(loaded, model) : undefined
+        const r = subagentModel(modelInfo(engine.model), wanted && modelInfo(wanted), engine.personSaid())
+        return "refused" in r ? { error: r.refused } : { model: r.model }
       } catch (e) {
         return { error: (e as Error).message }
       }
@@ -719,6 +772,14 @@ export function createApp(o: AppOptions): App {
   })
   // The lead's turn ended: it is idle, and its slot frees (team.md §8).
   bus.on((e) => e.type === "done" && teamRt.leadDone())
+  // Spec 3.4.0: the main agent's mode reaches down the tree; its progress and context are in it.
+  bus.on((e) => {
+    if (e.type === "mode") teamRt.mainMode(e.mode)
+    else if (e.type === "phase") {
+      teamRt.team.setProgress(TEAM.names.lead, e.progress)
+      if (teamRt.team.hasMembers()) teamRt.notify()
+    } else if (e.type === "usage" && !e.child && e.context) teamRt.team.setContext(TEAM.names.lead, e.used ?? e.usage.input + e.usage.output, e.context)
+  })
 
   // MCP tools join the engine's list as servers connect (the engine reads it at each prompt).
   let mcpNames = new Set<string>()

src/bus/index.ts

@@ -47,7 +47,7 @@ export type Event =
   | { type: "steered"; texts: string[]; ids?: string[] }
   /** A task — one prompt, with every step and tool it takes — starts or ends; `turnId` as on `prompt`. */
   | { type: "task"; state: "start" | "end"; turnId?: string }
-  | { type: "notice"; message: string; level?: "info" | "warn" }
+  | { type: "notice"; message: string; level?: "info" | "warn"; /** Spec 3.4.0: a request that will be sent again. */ retry?: { reason: string; in: number; attempt?: number } }
   /** Discard the reply streamed so far in this step: it is being retried. */
   | { type: "retract" }
   /** The session's plan was written, changed or cleared (plan_write, plan_update, /new). */
@@ -65,6 +65,9 @@ export type Event =
   /** The session's team changed (spec 3.2.0): a member added, started, finished or stopped, a
    *  message in the room. Read the team itself for what. */
   | { type: "team" }
+  /** Spec 3.4.0 (harness/progress.md): what the session's agent is doing now, or null. A
+   *  subagent's progress is in the tree (`team`). */
+  | { type: "phase"; progress: import("../session/progress.ts").Progress | null }
   /** The permission mode changed from inside a turn (plan approved). */
   | { type: "mode"; mode: Mode }
   /** The session's model or effort changed, for whatever reason — /model, /effort, the settings
@@ -127,7 +130,7 @@ export interface Asker {
     purpose?: string
     /** Set when a subagent running as a background job asks, so the card can say which — it may
      *  ask while the session itself is idle. */
-    source?: { jobId: string; agent: string; description: string }
+    source?: { jobId: string; agent: string; description: string; /** Spec 3.4.0: a subagent of the tree. */ subagent?: { id: string; name: string; position: string } }
   }): Promise<AskReply>
   /** The ask_user tool. Absent: the tool reports that nobody can be asked. `callId`: the tool call's
    *  id, which keys the card on another screen. */

src/cli.ts

@@ -98,7 +98,7 @@ function fail(msg: string, code = 1): never {
 }
 
 async function loginCommand(cmd: "login" | "logout", args: string[]) {
-  const { codeEvent, codeInstructions, doneEvent, instances, isWebuiEntry, login, loginSummary, loginWithToken, logout, sync, LembasError } = await import("./lembas/login.ts")
+  const { codeEvent, codeInstructions, doneEvent, instances, isWebuiEntry, login, loginSummary, loginWithToken, logout, sync, LembasError, SignInRejected } = await import("./lembas/login.ts")
   const { values, positionals } = parseArgs({
     args,
     allowPositionals: true,
@@ -144,10 +144,19 @@ async function loginCommand(cmd: "login" | "logout", args: string[]) {
     // `login` again with no address refreshes the one instance there is.
     const again = address ? Object.values(known).find((i) => sameBase(i.base_url, address)) : Object.keys(known).length === 1 ? Object.values(known)[0] : undefined
     if (again && !values.ca && isWebuiEntry(again.connection)) {
+      try {
         const r = await sync(again.connection, { setDefault: values.default })
         if (json) emit(doneEvent(r, false))
         else console.log(loginSummary(r, false).join("\n"))
         return
+      } catch (e) {
+        // The token here was revoked or is unknown to the instance: refreshing it can never work,
+        // and "run login again" would send the person straight back here. Sign in afresh instead;
+        // the new login replaces the old one.
+        if (!(e instanceof SignInRejected)) throw e
+        if (!json) console.error("The instance no longer accepts the sign-in kept here: signing in again.")
+        address = again.base_url
+      }
     }
     // An older kind of login (the models copied into connections.yaml) is replaced by signing in
     // again, at the address it had.

src/config/schema.ts

@@ -60,8 +60,8 @@ export const ModelSpec = z.strictObject({
   headers: z.record(z.string(), z.string()).optional().describe("Extra HTTP headers for this model's requests."),
   body: z.record(z.string(), z.unknown()).optional().describe("Merged into every request body for this model."),
   notes: z.string().max(300).optional().describe("What this model is good at, in a line: the model in use is told which others there are, for handing work over (task) or switching."),
-  single_session: z.boolean().optional().describe("The server serves one request at a time: a subagent on this same model would take its only slot and push the session's cached prompt out, so it is refused. The same as parallel: 1."),
+  single_session: z.boolean().optional().describe("The server serves one request at a time: a subagent on this same model would take its only slot and push the session's cached prompt out, so it is refused (name a model on another server instead). The same as parallel: 1."),
-  parallel: z.number().int().positive().optional().describe("How many requests the model's server serves at once (capacity.parallel); wins over single_session. Absent with no single_session: no limit (a hosted API). A team is offered only on a model that serves at least 2."),
+  parallel: z.number().int().positive().optional().describe("How many requests the model's server serves at once (capacity.parallel); wins over single_session. Absent with no single_session: no limit (a hosted API). The session's model is its own subagents' model only when it serves at least 2."),
   group: z.string().optional().describe("Models sharing one GPU server that holds one model at a time (llama-swap), as the LLeMbas server says (capacity.group): a subagent on another model of the same group would unload this one, so it is refused."),
 })
 export type ModelSpec = z.infer<typeof ModelSpec>
@@ -298,12 +298,12 @@ export const Config = z.strictObject({
   effort: z.union([z.enum(EFFORTS), z.literal("off"), z.literal("default")]).optional().describe("The effort level a session starts with, when the model offers it: an effort, off (thinking switched off, on a model with the switch) or default; otherwise the model's own start level."),
   agents: z
     .strictObject({
-      min: z.number().int().min(1).optional().describe("The fewest agents a team works with, the lead included (default 1). 2 or more: the lead works through a team from the start."),
+      min: z.number().int().min(1).optional().describe("Subagents, Min: the agents at work at least, this agent included (default 1 — 1 is this agent alone). 2 or more: it works through subagents from the start."),
-      max: z.number().int().min(1).optional().describe("The most agents a team may have, the lead included (default 4). 1: one agent, no team and no helper."),
+      max: z.number().int().min(1).optional().describe("Subagents, Max: the agents at work at most, this agent included (default 4 — this agent and up to 3 subagents). 1: this agent alone, no subagents."),
       cap: z.number().int().min(1).optional().describe("Global config only. The highest max a session may set (default 32)."),
     })
     .optional()
-    .describe("Teams (harness team.md): how many agents a session's team may have. Offered only where subagents are on and the model serves two requests at once. /agents changes it for the session."),
+    .describe("Subagents (harness team.md, subagents.md): how many agents a session may have at work, counting the session's own agent — Subagents: Min 1, Max 4 by default; 1 is this agent alone. Offered where subagents are on and a model could serve one: the session's, when it serves two requests at once, or another model off its one-model server. /team size and /effort (one range, Min and Max) change it for the session."),
   subagents: z.boolean().optional().describe("false: no subagents — no task tool and no team."),
   settings_tool: z.enum(["ask", "allow", "off"]).optional().describe("Global config only. The agent's settings tool: ask (default) — each change is approved like any tool call; allow — changes go through without asking (a less strict mode still asks); off — no settings tool."),
   update: Update.optional().describe("Global config only. Updating LLeMbas CLI itself."),

src/config/settings.ts

@@ -9,6 +9,7 @@ import { existsSync, readFileSync } from "node:fs"
 import { join } from "node:path"
 import { parse, parseDocument } from "yaml"
 import { z } from "zod"
+import LOOP_JSON from "../../harness/loop.json"
 import { paths } from "./paths.ts"
 import { asMode, Config, EFFORTS, MODES, PERSONALITIES, UPDATE_CHANNELS } from "./schema.ts"
 
@@ -29,8 +30,13 @@ export interface SettingDef {
   reload?: boolean
   /** What it is when nothing sets it. */
   fallback?: string
+  /** What a person reads beside the key in /settings. */
+  hint?: string
 }
 
+// The Subagents setting's names and what its numbers count (harness loop.json subagents.setting).
+const SUB = LOOP_JSON.subagents.setting
+
 const ALL: Scope[] = ["session", "global", "project"]
 const USER: Scope[] = ["session", "global"]
 
@@ -38,8 +44,8 @@ export const SETTINGS: SettingDef[] = [
   { key: "model", kind: "model", agent: true, scopes: ALL },
   { key: "effort", kind: "effort", agent: true, scopes: ALL, fallback: "the model's start level" },
   // The team's size is the person's: the agent never sets how many agents it may have.
-  { key: "agents.min", kind: "number", agent: false, scopes: ALL, fallback: "1" },
+  { key: "agents.min", kind: "number", agent: false, scopes: ALL, fallback: "1", hint: `${SUB.label}: ${SUB.min} — ${SUB.short}` },
-  { key: "agents.max", kind: "number", agent: false, scopes: ALL, fallback: "4" },
+  { key: "agents.max", kind: "number", agent: false, scopes: ALL, fallback: "4", hint: `${SUB.label}: ${SUB.max} — this agent included; 4 is this agent and up to 3 subagents` },
   { key: "mode", kind: "enum", values: MODES, agent: true, scopes: ALL, fallback: "manual" },
   { key: "small_model", kind: "model", agent: true, scopes: ALL, fallback: "the session's model" },
   { key: "knowledge", kind: "list", agent: true, scopes: ALL, fallback: "every base" },

src/harness.ts

@@ -52,6 +52,13 @@ import t_team_add from "../harness/tools/team_add.json"
 import t_team_read from "../harness/tools/team_read.json"
 import t_team_send from "../harness/tools/team_send.json"
 import t_team_stop from "../harness/tools/team_stop.json"
+import t_team_assign from "../harness/tools/team_assign.json"
+import t_team_decide from "../harness/tools/team_decide.json"
+import t_team_findings from "../harness/tools/team_findings.json"
+import t_team_finish from "../harness/tools/team_finish.json"
+import t_team_replace from "../harness/tools/team_replace.json"
+import t_team_request from "../harness/tools/team_request.json"
+import t_team_whoami from "../harness/tools/team_whoami.json"
 import t_view_image from "../harness/tools/view_image.json"
 import t_web_fetch from "../harness/tools/web_fetch.json"
 import t_web_search from "../harness/tools/web_search.json"
@@ -76,8 +83,9 @@ export interface ToolSpecDef {
   purpose: string | false
   /** Names each project used before spec v1; a call under one is read as this tool. */
   aliases: { cli: string[]; llembas: string[] }
-  /** `subagent`: offered only to a subagent (report), never to a session's own model. */
+  /** `subagent`: offered only to a subagent (report, team_finish, team_request), never to a
-  offered?: "subagent"
+   *  session's own model; `main` (spec 3.4.0): only to the main agent, never to a subagent. */
+  offered?: "subagent" | "main"
   argument_aliases?: Record<string, string>
   /** Spec 3.0.0: what must hold for it to be offered — `display` (the session is shown where a file
    *  can be opened: over the link, to a client that offers `show`), `page`, `forge` (the chat's);
@@ -118,16 +126,62 @@ export const LOOP: {
     running_max: { default: number }
     subagent_steps: { default: number }
   }
+  /** Subagents (spec 3.3.0, harness/subagents.md): the model refusals, what names a model, and how
+   *  the size setting is shown. */
+  subagents: {
+    texts: { not_named: string; same_group: string; one_request: string }
+    named: { before: string; after: string }
+    setting: { label: string; min: string; max: string; hint: string; short: string; own: string }
+  }
   /** Teams (spec 3.2.0, harness/team.md): the size setting, names, marks, texts. */
   team: {
     agents: { min: number; max: number; cap: number }
-    names: { pattern: string; lead: string; reserved: string[] }
+    names: { pattern: string; lead: string; reserved: string[]; parent: string; id_prefix: string }
-    marks: { message: string; report: string }
+    marks: { message: string; report: string; assignment: string; request: string; decision: string; context_full: string }
     unavailable: { one_agent: string; one_request: string }
     work: { write: string[]; execute: string[]; kept: string[] }
     read_max: number
-    texts: Record<"added" | "added_waiting" | "sent" | "stopped" | "manager" | "full" | "bad_name" | "all_name" | "taken" | "no_goal" | "unknown" | "self" | "nobody" | "not_yours" | "nobody_to_stop" | "no_report" | "ended", string>
+    texts: Record<
-    read: { tree: string; agent: string; lead_role: string; states: Record<"working" | "waiting" | "finished" | "idle", string>; active: string; room: string; message: string; report: string; none: string; more: string }
+      | "added" | "added_waiting" | "sent" | "stopped" | "manager" | "full" | "bad_name" | "all_name" | "taken" | "no_goal" | "unknown" | "self" | "nobody" | "not_yours" | "nobody_to_stop" | "no_report" | "ended"
+      // Spec 3.4.0 (subagents.md §5–§13).
+      | "only_main" | "not_manager" | "wider" | "request_not_narrower" | "request_kind" | "requestable" | "not_delegated" | "not_requestable" | "requested" | "no_request" | "decided" | "allowed" | "refused" | "decision_allowed" | "decision_refused" | "no_parent" | "finished" | "has_ended" | "cycle" | "assigned" | "replaced" | "replace_brief" | "no_findings" | "noted" | "not_yours_findings" | "context_full" | "gone" | "request_body",
+      string
+    >
+    read: {
+      tree: string
+      agent: string
+      lead_role: string
+      states: Record<"working" | "waiting" | "finished" | "idle" | "paused" | "done" | "out_of_context" | "stopped", string>
+      active: string
+      room: string
+      message: string
+      report: string
+      none: string
+      more: string
+      slots: string
+      ended: string
+      ended_agent: string
+      context: string
+      request: string
+    }
+    /** Spec 3.4.0. */
+    main_only: { tools: string[] }
+    context: { pause_at: number; final_steps: number }
+    request: { kinds: string[]; id_prefix: string }
+    whoami: { lines: Record<"who" | "parent" | "task" | "mode" | "request" | "model" | "limits" | "spawn_yes" | "spawn_no" | "context", string> }
+    findings: { header: string; item: string }
+  }
+  /** Progress (spec 3.4.0, harness/progress.md): the phases, their texts, the line. */
+  progress: {
+    phases: string[]
+    texts: Record<string, string>
+    reason: Record<string, string>
+    what: Record<string, string>
+    stopped_reason: Record<string, string>
+    line: string
+    no_elapsed: string[]
+    what_max: number
+    pace_per_second: number
   }
   /** Context hygiene (spec 3.0.0, harness/context.md): the result ids, the texts of a read's
    *  header, a stub, the elided line, recall; the output cap; what is deduplicated; the table
@@ -165,6 +219,8 @@ export const LOOP: {
       team_lead: string
       team_one: string
       team_stopped: string
+      /** Spec 3.3.0 (subagents.md §4): the session's model cannot serve subagents itself. */
+      team_own: string
     }
     /** How long a session waits for its configured MCP servers before its first request. */
     mcp_wait_ms: number
@@ -245,7 +301,7 @@ export function narrowParameters(parameters: Record<string, unknown>, narrow: Na
 }
 
 export const TOOL_SPECS: Record<string, ToolSpecDef> = Object.fromEntries(
-  ([t_apply_patch, t_ask_user, t_bash, t_branch, t_commit, t_decisions, t_edit, t_glob, t_grep, t_job_list, t_job_output, t_job_send, t_job_stop, t_knowledge_get, t_knowledge_search, t_list, t_memory, t_merge, t_mkdir, t_multiedit, t_mv, t_note_manage, t_note_view, t_notes_search, t_plan_submit, t_plan_update, t_plan_write, t_pr, t_python, t_read, t_recall, t_release, t_report, t_rm, t_search, t_session_search, t_settings, t_show, t_skill_manage, t_skill_view, t_skills_list, t_tag, t_task, t_tasks, t_team_add, t_team_read, t_team_send, t_team_stop, t_view_image, t_web_fetch, t_web_search, t_write] as ToolSpecDef[]).map((t) => [t.name, t]),
+  ([t_apply_patch, t_ask_user, t_bash, t_branch, t_commit, t_decisions, t_edit, t_glob, t_grep, t_job_list, t_job_output, t_job_send, t_job_stop, t_knowledge_get, t_knowledge_search, t_list, t_memory, t_merge, t_mkdir, t_multiedit, t_mv, t_note_manage, t_note_view, t_notes_search, t_plan_submit, t_plan_update, t_plan_write, t_pr, t_python, t_read, t_recall, t_release, t_report, t_rm, t_search, t_session_search, t_settings, t_show, t_skill_manage, t_skill_view, t_skills_list, t_tag, t_task, t_tasks, t_team_add, t_team_assign, t_team_decide, t_team_findings, t_team_finish, t_team_read, t_team_replace, t_team_request, t_team_send, t_team_stop, t_team_whoami, t_view_image, t_web_fetch, t_web_search, t_write] as ToolSpecDef[]).map((t) => [t.name, t]),
 )
 
 /** Whether the spec gives a tool to this CLI: not the server's alone (`scope: llembas`), and

src/lembas/client.ts

@@ -27,6 +27,8 @@ export function protocolMismatch(base: string, d: Pick<Discovery, "protocol" | "
 }
 
 export class LembasError extends Error {}
+/** The instance answered 401: the token this machine holds is revoked or unknown there. `login` signs in afresh. */
+export class SignInRejected extends LembasError {}
 /** The server and this CLI speak no link protocol in common: no login, and no dial gets further. */
 export class ProtocolError extends LembasError {}
 
@@ -268,7 +270,7 @@ export async function checkToken(d: Discovery, token: string, tls: Tls | undefin
 
 export async function models(d: { api: { openai: string } }, token: string, tls: Tls | undefined, timeoutMs?: number): Promise<ServedModel[]> {
   const res = await call(`${d.api.openai}/models`, { headers: { authorization: `Bearer ${token}`, accept: "application/json" } }, tls, timeoutMs)
-  if (res.status === 401) throw new LembasError("the instance no longer accepts this sign-in — run login again")
+  if (res.status === 401) throw new SignInRejected("the instance no longer accepts this sign-in — run login again")
   if (res.status === 403) throw new LembasError("this account may not use the instance's models through the API (the “Use the API” permission)")
   if (!res.ok) throw new LembasError(`listing the models: ${res.status} ${res.statusText}`)
   const body = (await res.json()) as { data?: ServedModel[] }
@@ -293,7 +295,7 @@ export interface InstanceInfo {
 export async function instanceInfo(base: string, token: string, tls: Tls | undefined, timeoutMs?: number): Promise<InstanceInfo | undefined> {
   const res = await call(`${base}/api/v1/instance`, { headers: { authorization: `Bearer ${token}`, accept: "application/json" } }, tls, timeoutMs)
   if (res.status === 404) return undefined
-  if (res.status === 401) throw new LembasError("the instance no longer accepts this sign-in — run login again")
+  if (res.status === 401) throw new SignInRejected("the instance no longer accepts this sign-in — run login again")
   if (!res.ok) throw new LembasError(`reading the instance: ${res.status} ${res.statusText}`)
   const body = (await res.json().catch(() => undefined)) as InstanceInfo | undefined
   return body && typeof body === "object" ? body : undefined
@@ -341,7 +343,7 @@ export interface Usage {
 export async function usage(base: string, token: string, tls: Tls | undefined, timeoutMs = 5000): Promise<Usage | undefined> {
   const res = await call(`${base}/v1/usage`, { headers: { authorization: `Bearer ${token}`, accept: "application/json" } }, tls, timeoutMs)
   if (res.status === 404) return undefined
-  if (res.status === 401) throw new LembasError("the instance no longer accepts this sign-in — run login again")
+  if (res.status === 401) throw new SignInRejected("the instance no longer accepts this sign-in — run login again")
   if (!res.ok) throw new LembasError(`reading the usage: ${res.status} ${res.statusText}`)
   const body = (await res.json().catch(() => undefined)) as Usage | undefined
   return body && typeof body === "object" ? body : undefined
@@ -369,7 +371,7 @@ function asPersonalization(b: unknown): Personalization | undefined {
 export async function personalization(base: string, token: string, tls: Tls | undefined, timeoutMs = 5000): Promise<Personalization | undefined> {
   const res = await call(`${base}/v1/me/personalization`, { headers: { authorization: `Bearer ${token}`, accept: "application/json" } }, tls, timeoutMs)
   if (res.status === 404) return undefined
-  if (res.status === 401) throw new LembasError("the instance no longer accepts this sign-in — run login again")
+  if (res.status === 401) throw new SignInRejected("the instance no longer accepts this sign-in — run login again")
   if (!res.ok) throw new LembasError(`reading the personalization: ${res.status} ${res.statusText}`)
   return asPersonalization(await res.json().catch(() => undefined))
 }
@@ -393,7 +395,7 @@ export async function savePersonalization(base: string, token: string, tls: Tls
     if (res.status === 422) throw new LembasError(`the instance would not take that personalization${detail ? `: ${detail}` : ""}`)
     throw new LembasError(detail ? `the instance did not allow it: ${detail}` : "not allowed: the instance refused to save the personalization")
   }
-  if (res.status === 401) throw new LembasError("the instance no longer accepts this sign-in — run login again")
+  if (res.status === 401) throw new SignInRejected("the instance no longer accepts this sign-in — run login again")
   if (!res.ok) throw new LembasError(`saving the personalization: ${res.status} ${res.statusText}`)
   return asPersonalization(await res.json().catch(() => undefined))
 }

src/lembas/login.ts

@@ -13,7 +13,7 @@ import { chmodSync, existsSync, mkdirSync, readFileSync, rmSync, writeFileSync }
 import { join, resolve } from "node:path"
 import { parseDocument } from "yaml"
 import { expandHome, globalPaths, isLocal, paths } from "../config/paths.ts"
-import { awaitToken, checkToken, discover, LembasError, signOut, spoken, startDevice, type DeviceStart, type Discovery, type Granted, type InstanceInfo, type ServedModel, type Tls } from "./client.ts"
+import { awaitToken, checkToken, discover, LembasError, SignInRejected, signOut, spoken, startDevice, type DeviceStart, type Discovery, type Granted, type InstanceInfo, type ServedModel, type Tls } from "./client.ts"
 import { dropCache, fetchWebui, isChatModel, readCache, specFor } from "./webui.ts"
 
 export interface Instance {
@@ -557,4 +557,4 @@ export async function logout(connection: string): Promise<LogoutResult> {
   return { connection, revoked, modelLeft: model && (ours || model.startsWith(`${connection}/`)) ? model : undefined }
 }
 
-export { LembasError, readCache, specFor }
+export { LembasError, readCache, SignInRejected, specFor }

src/project/agents.ts

@@ -13,6 +13,7 @@ export interface AgentSpec {
   name: string
   description: string
   instructions: string
+  /** Read, and not used for a subagent (spec 3.3.0, subagents.md §2): shown so a file that sets it says so. */
   model?: string
   /** A fixed permission mode; absent = the parent's. */
   mode?: Mode

src/prompt/assemble.ts

@@ -20,6 +20,7 @@ import blockSubagent from "../../harness/prompts/blocks/subagent.md" with { type
 import blockContext from "../../harness/prompts/blocks/context.md" with { type: "text" }
 import blockTeamLead from "../../harness/prompts/blocks/team_lead.md" with { type: "text" }
 import blockTeamMember from "../../harness/prompts/blocks/team_member.md" with { type: "text" }
+import blockTeamSubmanager from "../../harness/prompts/blocks/team_submanager.md" with { type: "text" }
 import agentExplore from "../../harness/prompts/agents/explore.md" with { type: "text" }
 import agentGeneral from "../../harness/prompts/agents/general.md" with { type: "text" }
 import modeManual from "../../harness/prompts/modes/manual.md" with { type: "text" }
@@ -61,6 +62,7 @@ const BUILTIN: Record<string, string> = {
   "blocks/context.md": blockContext,
   "blocks/team_lead.md": blockTeamLead,
   "blocks/team_member.md": blockTeamMember,
+  "blocks/team_submanager.md": blockTeamSubmanager,
   "agents/explore.md": agentExplore,
   "agents/general.md": agentGeneral,
   "modes/manual.md": modeManual,
@@ -236,9 +238,9 @@ export interface PromptContext {
   /** Set for a subagent: who it is and what it is for; replaces the mode's own instructions.
    *  `report`: it has the report tool. */
   subagent?: { name: string; instructions: string; report?: boolean }
-  /** Spec 3.2.0 (team.md §4): a team member — blocks/team_member.md in place of blocks/subagent.md,
+  /** Spec 3.2.0 (team.md §4), 3.4.0: a subagent of the tree — blocks/team_member.md in place of
-   *  with the whole of blocks/team_lead.md for one that may lead. */
+   *  blocks/subagent.md, with blocks/team_submanager.md for a sub-manager. */
-  member?: { name: string; manager: string; role: string; goal: string; lead: boolean }
+  member?: { id: string; name: string; manager: string; role: string; goal: string; lead: boolean }
   /** Spec 3.2.0 (team.md §10): the team tools are offered to this session's agent (the lead):
    *  blocks/team_lead.md is in its prompt. */
   teamLead?: boolean
@@ -345,7 +347,7 @@ export function assembleSystem(ctx: PromptContext): string {
     ].join("\n"),
   )
   if (ctx.roster && !ctx.subagent)
-    sections.push(`Other models configured here — the task tool can run a subagent on one (its model argument); the user can switch to one, and so can you through the settings tool when they ask:\n<models>\n${ctx.roster}\n</models>`)
+    sections.push(`Other models configured here — a subagent runs on one (task's or team_add's model argument) only when the user names it; the user can switch to one, and so can you through the settings tool when they ask:\n<models>\n${ctx.roster}\n</models>`)
   if (ctx.git) sections.push(`The git repository as it was when this session started — a snapshot, not kept up to date; run git status for how it is now:\n<git>\n${ctx.git}\n</git>`)
   if (ctx.tasks && !ctx.subagent) sections.push(`Open items on the project's task board (the tasks tool changes it):\n<project_tasks>\n${ctx.tasks}\n</project_tasks>`)
 
@@ -391,12 +393,14 @@ export function assembleSystem(ctx: PromptContext): string {
     sections.push(
       fillPrompt(promptText("blocks/team_member.md"), {
         name: ctx.member.name,
+        id: ctx.member.id,
         manager: ctx.member.manager,
         role: ctx.member.role,
         goal: ctx.member.goal,
         instructions: ctx.subagent.instructions,
         read_only: ctx.mode === "plan" ? SUBAGENT_READ_ONLY : "",
-        lead: ctx.member.lead ? promptText("blocks/team_lead.md") : "",
+        // Spec 3.4.0: a sub-manager directs; it no longer gets the whole lead block.
+        lead: ctx.member.lead ? promptText("blocks/team_submanager.md") : "",
       })
         .replace(/\n{3,}/g, "\n\n")
         .trim(),

src/provider/common.ts

@@ -54,7 +54,8 @@ export async function* retrying(attempt: () => AsyncGenerator<StreamEvent>, hand
       // Twice: the first at once, the second after a pause (a model still loading, a busy server).
       if ((e.status ?? 0) >= 500 && retried5xx < 2) {
         retried5xx++
-        yield { type: "notice", message: `server error, retrying (${retried5xx} of 2) — ${e.message.split("\n")[0]!.slice(0, 160)}` }
+        const reason = e.message.split("\n")[0]!.slice(0, 160)
+        yield { type: "notice", message: `server error, retrying (${retried5xx} of 2) — ${reason}`, retry: { reason, in: retried5xx === 2 ? 2 : 0, attempt: retried5xx } }
         if (retried5xx === 2) await new Promise((r) => setTimeout(r, 2000))
         continue
       }

src/provider/types.ts

@@ -54,7 +54,7 @@ export type StreamEvent =
   | { type: "progress"; total: number; cache: number; processed: number; ms: number }
   | { type: "finish"; reason: string; message: Extract<Message, { role: "assistant" }> }
   /** Something worth telling the user that is not an error, e.g. a retry. */
-  | { type: "notice"; message: string }
+  | { type: "notice"; message: string; retry?: { reason: string; in: number; attempt?: number } }
 
 export interface ChatRequest {
   system: string

src/session/capacity.ts

@@ -20,21 +20,21 @@ export function parallelOf(m: Pick<ResolvedModel, "spec">): number {
   return m.spec.single_session ? 1 : Infinity
 }
 
-/** `busy`: this session's requests the helper's model already serves besides the main prompt's
+/** `busy`: this session's requests the sub's model already serves besides the main prompt's
  *  (subagents running on it). */
-export function capacityRefusal(main: ResolvedModel, helper: ResolvedModel, busy = 0): string {
+export function capacityRefusal(main: ResolvedModel, sub: ResolvedModel, busy = 0): string {
-  if (helper.ref === main.ref) {
+  if (sub.ref === main.ref) {
-    const n = parallelOf(helper)
+    const n = parallelOf(sub)
     if (n === 1)
-      return `${helper.ref} serves one request at a time, so it cannot be its own subagent: the subagent would take the server's only slot and push this session's cached prompt out of it.`
+      return `${sub.ref} serves one request at a time, so it cannot be its own subagent: the subagent would take the server's only slot and push this session's cached prompt out of it.`
     if (1 + busy >= n)
-      return `${helper.ref} serves ${n} requests at once, and this session already has ${n} on it; another subagent would wait, or push this session's cached prompt out.`
+      return `${sub.ref} serves ${n} requests at once, and this session already has ${n} on it; another subagent would wait, or push this session's cached prompt out.`
     return ""
   }
   const shared = groupOf(main)
-  if (shared !== undefined && shared === groupOf(helper)) {
+  if (shared !== undefined && shared === groupOf(sub)) {
     const where = `${main.ref}'s server (group ${main.spec.group})`
-    return `${where} holds one model at a time, so a subagent on ${helper.ref} would unload ${main.ref}, and this session's cached prompt with it, while its turn waits.`
+    return `${where} holds one model at a time, so a subagent on ${sub.ref} would unload ${main.ref}, and this session's cached prompt with it, while its turn waits.`
   }
   return ""
 }
@@ -42,17 +42,17 @@ export function capacityRefusal(main: ResolvedModel, helper: ResolvedModel, busy
 /** The same rule between two subagents: a new one is not started where its model would unload, or
  *  take the only slot of, a subagent already running (the spec's "beside the model of a subagent
  *  already running"). */
-export function capacityBeside(running: ResolvedModel, helper: ResolvedModel, busy = 1): string {
+export function capacityBeside(running: ResolvedModel, sub: ResolvedModel, busy = 1): string {
-  if (helper.ref === running.ref) {
+  if (sub.ref === running.ref) {
-    const n = parallelOf(helper)
+    const n = parallelOf(sub)
-    if (n === 1) return `${helper.ref} serves one request at a time, and a subagent is already running on it; a second would wait for the first, or push it out.`
+    if (n === 1) return `${sub.ref} serves one request at a time, and a subagent is already running on it; a second would wait for the first, or push it out.`
-    if (busy >= n) return `${helper.ref} serves ${n} requests at once, and as many subagents already run on it; another would wait, or push one out.`
+    if (busy >= n) return `${sub.ref} serves ${n} requests at once, and as many subagents already run on it; another would wait, or push one out.`
     return ""
   }
   const shared = groupOf(running)
-  if (shared !== undefined && shared === groupOf(helper)) {
+  if (shared !== undefined && shared === groupOf(sub)) {
     const where = `${running.ref}'s server (group ${running.spec.group})`
-    return `${where} holds one model at a time, and a subagent is already running on ${running.ref} there; one on ${helper.ref} would unload it.`
+    return `${where} holds one model at a time, and a subagent is already running on ${running.ref} there; one on ${sub.ref} would unload it.`
   }
   return ""
 }

src/session/commands.ts

@@ -17,6 +17,9 @@ import { fill, promptText } from "../prompt/assemble.ts"
 import { skillMessage, slug, type Skill } from "../skill/index.ts"
 import { CONTINUE_PROMPT } from "./engine.ts"
 import { TEAM, Team } from "../team/team.ts"
+import { SUBAGENTS } from "../team/subagents.ts"
+import { progressLine } from "./progress.ts"
+import { fillPrompt } from "../prompt/fill.ts"
 
 export type CommandKind = "custom" | "skill" | "builtin"
 
@@ -48,7 +51,7 @@ export const SHARED_BUILTINS: { name: string; description: string; hint?: string
   { name: "init", description: "make or update the project's AGENTS.md from what the repository says", hint: "what to focus on" },
   { name: "continue", description: "carry on — after esc stopped it, or when it should keep going (ctrl+g)" },
   { name: "agents", description: "subagents: list them, make one, or hand one a task", hint: "name" },
-  { name: "team", description: "the team: its agents and the room; its size; stop a member", hint: "size <min> <max> | stop <name>" },
+  { name: "team", description: "subagents: the tree and the room; how many; stop one; their findings", hint: "size <min> <max> | stop <name> | findings <name>" },
 ]
 
 /** The status bar's word for a team with members: its size of the most it may have, and how many
@@ -56,7 +59,35 @@ export const SHARED_BUILTINS: { name: string; description: string; hint?: string
 export function teamLabel(team: Team): string {
   if (!team.hasMembers()) return ""
   const n = team.active()
-  return `team ${team.size()}/${team.max}${n ? ` · ${n} working` : ""}`
+  return `agents ${team.size()}/${team.max}${n ? ` · ${n} subagent${n === 1 ? "" : "s"} working` : ""}`
+}
+
+/** Spec 3.4.0 (subagents.md §12): the tree as the person sees it — every agent with its position,
+ *  state, model, mode, context use, progress line and messages; open slots; the ended ones. */
+export function treeText(team: Team, now = Date.now()): string {
+  const info = team.info()
+  const R = TEAM.read
+  const depth = (a: { parent: string | null }): number => (a.parent ? 1 + depth(info.agents.find((x) => x.name === a.parent) ?? { parent: null }) : 0)
+  const lines = [`${SUBAGENTS.setting.label} (${info.size} of ${info.max} agents; ${SUBAGENTS.setting.min} ${info.min}, ${SUBAGENTS.setting.max} ${info.max}):`]
+  for (const a of info.agents) {
+    const bits = [
+      `${"  ".repeat(depth(a))}- ${a.name}${a.id === a.name ? "" : ` (${a.id})`} — ${a.position}`,
+      R.states[a.state],
+      a.model,
+      a.request.length ? `${a.mode} + ${a.request.join(", ")}` : a.mode,
+      a.context && a.context.window ? `${Math.floor((a.context.used * 100) / a.context.window)}% context` : "",
+      a.progress ? progressLine(a.progress, now) : "",
+      a.messages ? `${a.messages} message${a.messages === 1 ? "" : "s"}` : "",
+      a.findings ? `${a.findings} finding${a.findings === 1 ? "" : "s"}` : "",
+    ].filter(Boolean)
+    lines.push(bits.join(" · "))
+  }
+  if (info.open_slots) lines.push(fillPrompt(R.slots, { n: String(info.open_slots) }))
+  if (info.ended.length) {
+    lines.push(R.ended)
+    for (const a of info.ended) lines.push(`- ${a.name} (${a.id}) — ${a.position} · ${R.states[a.state]}${a.findings ? ` · ${a.findings} finding${a.findings === 1 ? "" : "s"}` : ""}`)
+  }
+  return lines.join("\n")
 }
 
 /** `/team [size <min> <max> | stop <name>]`: the team's tree and room, as the lead sees them (the
@@ -65,10 +96,11 @@ export function teamLabel(team: Team): string {
 export function teamText(app: App, arg: string): string {
   const rt = app.team
   const t = rt.team
+  const S = SUBAGENTS.setting
   const [verb, ...rest] = arg.trim().split(/\s+/).filter(Boolean)
   if (verb === "size") {
     const [min, max] = rest.map(Number)
-    if (min === undefined || max === undefined || Number.isNaN(min) || Number.isNaN(max)) return `/team size <min> <max> — now ${t.min} to ${t.max} (at most ${t.cap})`
+    if (min === undefined || max === undefined || Number.isNaN(min) || Number.isNaN(max)) return `/team size <min> <max> — now ${S.label}: ${S.min} ${t.min}, ${S.max} ${t.max} (Max at most ${t.cap}). ${S.hint}`
     // Checked whole before anything is set: a refused pair changes nothing (team.md §2).
     const no = Team.check(min, max, t.cap)
     if (no) return no
@@ -79,7 +111,12 @@ export function teamText(app: App, arg: string): string {
     } catch (e) {
       return (e as Error).message
     }
-    return `The team may have ${t.min} to ${t.max} agents for this session; the lead is told with your next message.`
+    return `${S.label}: ${S.min} ${t.min}, ${S.max} ${t.max} — ${S.short}. Set for this session; the agent is told with your next message.`
+  }
+  if (verb === "findings") {
+    const name = rest[0]
+    if (!name) return "/team findings <name or id>"
+    return t.readFindings("lead", name)
   }
   if (verb === "stop") {
     const name = rest[0]
@@ -88,9 +125,15 @@ export function teamText(app: App, arg: string): string {
     return gone.length ? `Stopped ${gone.join(", ")}; the lead is told with your next message.` : `No member called ${name} on the team.`
   }
   if (!t.offered)
-    return t.subagents ? `No team in this session: ${app.engine.model.ref} serves one request at a time, and a team needs two at once.` : "No team in this session: subagents are switched off (subagents: false)."
+    return t.subagents
+      ? `No subagents in this session: ${app.engine.model.ref} serves one request at a time, and every other model is on its one-model server.`
+      : "No subagents in this session: they are switched off (subagents: false)."
   const since = Math.max(0, t.room.length - TEAM.read_max)
-  return `${t.read("lead", since)}\n\nSize ${t.min} to ${t.max} agents (/team size <min> <max>); /team stop <name> stops a member.`
+  const own = t.couldServe().own ? "" : `\n${S.own}`
+  // The tree as the person sees it (spec 3.4.0 §12), then the room as team_read shows it.
+  const read = t.read("lead", since)
+  const room = read.slice(read.indexOf("\n\n") + 2)
+  return `${treeText(t)}\n\n${room}\n\n${S.label}: ${S.min} ${t.min}, ${S.max} ${t.max} — ${S.short} (/team size <min> <max>); /team stop <name> stops a subagent; /team open <name> shows its session, read-only; /team findings <name>; /context <name>.${own}`
 }
 
 /** `/agents [name]`: the subagents the task tool offers, or one of them — the text both the TUI and
@@ -100,7 +143,7 @@ export function agentsText(app: App, arg: string): string {
   const describe = (a: (typeof list)[number]) =>
     [
       `${a.name} (${a.source}) — ${a.description}`,
-      `  model ${a.model ?? "the session's"} · mode ${a.mode ?? "the session's"} · tools ${a.tools?.join(", ") ?? "all a subagent may have"}${a.isolation ? " · own worktree" : ""}`,
+      `  model the session's, or one you name${a.model ? ` (its own model: ${a.model} is not used)` : ""} · mode ${a.mode ?? "the session's"} · tools ${a.tools?.join(", ") ?? "all a subagent may have"}${a.isolation ? " · own worktree" : ""}`,
     ].join("\n")
   const name = arg.trim().split(/\s+/)[0]
   if (name) {

src/session/context.ts

@@ -41,23 +41,49 @@ export function contextBreakdown(engine: Engine, system: string, instructions: {
   if (imageCount) rows.push({ label: `images (${imageCount}, rough)`, tokens: by.images })
   const estimated = rows.filter((r) => !r.indent).reduce((n, r) => n + r.tokens, 0)
   const reported = engine.usage ? engine.contextUsed() : undefined
-  return { rows, estimated, reported, window: contextWindow(engine.model), model: engine.model.ref, last: engine.lastRequest }
+  // Where automatic compaction starts (EngineOptions compaction.autoAt): the reference's "N until
+  // auto-compact". None without a window, or when it is switched off (1 or more).
+  const autoAt = engine.o.compaction?.autoAt ?? 0.85
+  return { rows, estimated, reported, window: contextWindow(engine.model), autoAt: autoAt < 1 ? autoAt : undefined, model: engine.model.ref, last: engine.lastRequest }
 }
 
-/** Plain text, for the transcript. */
+/** One glyph per category in the segmented bar, in the order the categories are listed. */
+const SEGMENTS = ["█", "▓", "▒", "░", "▞", "▚", "▙", "▟"]
+
+/** Plain text, for the transcript: what is used of the window and how much, one segmented bar of
+ *  every category, each category with its tokens and its share of the window, and how far it is to
+ *  automatic compaction (/compact compacts now). */
 export function renderBreakdown(b: ReturnType<typeof contextBreakdown>): string {
-  const width = 24
+  const width = 40
-  const total = b.window ?? Math.max(b.estimated, b.reported ?? 0)
-  const bar = (n: number) => {
-    const cells = total ? Math.max(n > 0 ? 1 : 0, Math.round((n / total) * width)) : 0
-    return "█".repeat(Math.min(width, cells)).padEnd(width, "·")
-  }
-  const fmt = (n: number) => (n >= 1000 ? `${(n / 1000).toFixed(1)}k` : String(n))
-  const lines = [`context — ${b.model}${b.window ? `, window ${fmt(b.window)}` : ", window unknown"}`]
-  for (const r of b.rows) lines.push(`${r.indent ? "   ↳ " : "  "}${r.label.padEnd(r.indent ? 30 : 33)} ${fmt(r.tokens).padStart(7)}  ${r.indent ? "" : bar(r.tokens)}`)
   const used = b.reported ?? b.estimated
-  lines.push(`  ${"in use".padEnd(33)} ${fmt(used).padStart(7)}  ${bar(used)}${b.window ? `  ${Math.round((used / b.window) * 100)}%` : ""}`)
+  const total = b.window ?? Math.max(b.estimated, used)
-  lines.push(b.reported !== undefined ? "  (the server's own count from the last reply, plus an estimate since)" : "  (estimated: ≈4 characters a token — no reply yet to count from)")
+  const fmt = (n: number) => (n >= 1000 ? `${(n / 1000).toFixed(1)}k` : String(n))
+  const pct = (n: number) => (b.window ? `${Math.round((n / b.window) * 100)}%` : "")
+  const cats = b.rows.filter((r) => !r.indent)
+  // The categories' shares of the bar, scaled to what is in use (the server's count when it gave one).
+  const scale = b.estimated ? used / b.estimated : 1
+  let bar = ""
+  cats.forEach((r, i) => {
+    const cells = total && r.tokens > 0 ? Math.max(1, Math.round(((r.tokens * scale) / total) * width)) : 0
+    bar += SEGMENTS[i % SEGMENTS.length]!.repeat(cells)
+  })
+  bar = bar.slice(0, width).padEnd(width, "·")
+  const lines = [
+    `context — ${b.model}: ${fmt(used)}${b.window ? ` / ${fmt(b.window)} tokens · ${pct(used)} used` : " tokens, window unknown"}`,
+    `  ${bar}`,
+  ]
+  let i = 0
+  for (const r of b.rows) {
+    const mark = r.indent ? "   ↳" : ` ${SEGMENTS[i++ % SEGMENTS.length]} `
+    lines.push(`${mark} ${r.label.padEnd(r.indent ? 30 : 31)} ${fmt(r.tokens).padStart(7)}  ${r.indent ? "" : pct(r.tokens * scale).padStart(4)}`.trimEnd())
+  }
+  if (b.window) lines.push(`  · ${"free".padEnd(31)} ${fmt(Math.max(0, b.window - used)).padStart(7)}  ${pct(Math.max(0, b.window - used)).padStart(4)}`)
+  lines.push(`  ${"in use".padEnd(33)} ${fmt(used).padStart(7)}  ${pct(used).padStart(4)}`.trimEnd())
+  if (b.window && b.autoAt !== undefined) {
+    const until = Math.round(b.window * b.autoAt) - used
+    lines.push(until > 0 ? `  ${fmt(until)} until auto-compact (at ${Math.round(b.autoAt * 100)}%) — /compact to compact now` : `  past auto-compact (${Math.round(b.autoAt * 100)}%): the next step compacts — /compact to compact now`)
+  } else lines.push(b.window ? "  automatic compaction is off — /compact to compact now" : "  no automatic compaction without a window — /compact to compact now")
+  lines.push(b.reported !== undefined ? "  (the server's own count from the last reply, plus an estimate since; the categories are estimates scaled to it)" : "  (estimated: ≈4 characters a token — no reply yet to count from)")
   // What the server reused of the last request (harness context.md §10 rule 6): a slow reply with
   // little of it cached was a re-read.
   const last = b.last?.usage

src/session/engine.ts

@@ -8,7 +8,7 @@ import { isAbsolute } from "node:path"
 import type { Asker, AttachmentInfo, Bus } from "../bus/index.ts"
 import type { Mode } from "../config/schema.ts"
 import type { Level } from "../provider/effort.ts"
-import { evaluate, shadowsDeny, type Context as PermContext, type Rule } from "../permission/evaluate.ts"
+import { evaluate, shadowsDeny, type Context as PermContext, type PermissionRequest, type Rule } from "../permission/evaluate.ts"
 import { ProviderError, type Client, type ImagePart, type Message, type ResolvedModel, type ToolCallPart, type Usage } from "../provider/types.ts"
 import type { Plan } from "../plan/plan.ts"
 import { LOOP } from "../harness.ts"
@@ -22,6 +22,9 @@ import { estimateTokens, estimateValue } from "../provider/tokens.ts"
 import { duration } from "../duration.ts"
 import { memoryResults, presentIds, ResultContext } from "./results.ts"
 import { render } from "../plan/plan.ts"
+import { personTexts } from "../team/subagents.ts"
+import { ProgressTracker } from "./progress.ts"
+import type { Verdict } from "../team/delegation.ts"
 import { compactAsk, isNote, NO_CHANGING, noteFor, offeredFor, PREFIX_META, summaryOf, today, toldAt, unavailable, type Changing, type Kept, type Now, type Requirement, type Told } from "./prefix.ts"
 
 export interface EngineOptions {
@@ -77,6 +80,15 @@ export interface EngineOptions {
     /** What the lead was told of the team is to be said again: a new prompt, or a rewind. */
     forget?(): void
     beforeTurn?(signal: AbortSignal): Promise<boolean>
+    /** Spec 3.4.0 (subagents.md §7), a subagent: who decides its call — itself, the main agent
+     *  (through a request), or the person. Replaces the plain evaluation. */
+    permit?(request: PermissionRequest, ctx: Omit<PermContext, "mode">): Verdict
+    /** Spec 3.4.0 §9, a subagent: its context reached `pause_at` of its window before a step.
+     *  "stop": its turn ends now (it is paused; its final turn follows); "go": carry on without
+     *  compacting (it is in its final turn). Undefined: not a subagent of the tree. */
+    contextFull?(used: number, window: number): "stop" | "go" | undefined
+    /** Spec 3.4.0, a subagent: it ended by its own call (team_finish): the turn stops after this step. */
+    ended?(): boolean
   }
 }
 
@@ -86,6 +98,8 @@ export interface PromptMeta {
   attachments?: AttachmentInfo[]
   /** The harness started this turn — a job's note — not the person. */
   machine?: true
+  /** Spec 3.4.0: the steps this turn may take, when fewer than the engine's (a subagent's final turn). */
+  maxSteps?: number
 }
 
 const MAX_PARALLEL = LOOP.parallel.max
@@ -186,7 +200,11 @@ export class Engine {
   /** The prompt the team's offer was last decided for. */
   private teamFixed: Kept | undefined
 
+  /** Spec 3.4.0 (harness/progress.md): what this agent is doing, said as `phase` events. */
+  readonly progress: ProgressTracker
+
   constructor(readonly o: EngineOptions) {
+    this.progress = new ProgressTracker(o.bus)
     this.mode = o.mode
     this._effort = o.effort
     this.model = o.model
@@ -463,8 +481,19 @@ export class Engine {
   /** Before a step: past compaction.auto_at of the window, compact. Old tool results are never
    *  pruned in place (harness context.md: nothing already sent is rewritten — it would throw the
    *  prompt cache away from that point); compaction is the one deliberate break. */
-  private async manageContext(signal: AbortSignal) {
+  private async manageContext(signal: AbortSignal): Promise<"stop" | void> {
     const window = contextWindow(this.model)
+    // Spec 3.4.0 §9: a subagent of the tree does not compact; at pause_at it is paused instead.
+    const full = this.o.team?.contextFull
+    if (full && window) {
+      const used = this.contextUsed()
+      if (used !== undefined && used >= window * LOOP.team.context.pause_at) {
+        const r = full(used, window)
+        if (r === "stop") return "stop"
+        if (r === "go") return
+      }
+      if (full(0, 0) !== undefined) return
+    }
     const autoAt = this.o.compaction?.autoAt ?? 0.85
     if (!window || autoAt >= 1) return
     const limit = window * autoAt
@@ -481,6 +510,7 @@ export class Engine {
     }
     this.autoCompactions++
     this.o.bus.emit({ type: "notice", message: `context at ${Math.round((used / window) * 100)}% of ${window}: compacting the conversation` })
+    this.progress.set({ phase: "compacting" })
     try {
       await this.summarise(signal)
       // The task is never left to the summary: the prompt that started this turn comes back word
@@ -507,11 +537,13 @@ export class Engine {
 
   /** Time spent waiting for the user does not count against the wall-clock budget: somebody who
    *  thinks for ten minutes about one command has not spent the agent's time. */
-  private async waiting<T>(p: Promise<T>): Promise<T> {
+  private async waiting<T>(p: Promise<T>, what: "approval" | "question" | "plan" = "approval", tool?: string): Promise<T> {
     const t0 = performance.now()
+    const back = this.progress.hold({ phase: "waiting_for_person", what, ...(tool ? { tool } : {}) })
     try {
       return await p
     } finally {
+      back()
       this.turnWaitedMs += performance.now() - t0
     }
   }
@@ -611,6 +643,17 @@ export class Engine {
     if (this.o.store && this.sessionId) this.lastRowId = this.o.store.append(this.sessionId, m)
   }
 
+  /** What the person typed in this session, oldest first (subagents.md §2) — compacted messages
+   *  included, as the store keeps them; never a note, a marked message or a machine turn. */
+  personSaid(): string[] {
+    const all = this.o.store && this.sessionId ? this.o.store.messages(this.sessionId) : this.messages
+    return [...this.saidBefore, ...personTexts(all, this.o.steerMark ?? STEER_MARK)]
+  }
+
+  /** Spec 3.4.0 (acp.md, *Subagents in an agent chat*): the person's earlier words in a chat this
+   *  device has not seen, as the client sent them (`said` on session/new). */
+  saidBefore: string[] = []
+
   /** `extra`: attachments (@files, images) and the like — sent to the model after the text, never shown as typed.
    *  `meta`: the turn's id and what was attached, said on the `prompt` event. */
   async prompt(text: string, extra: (string | ImagePart)[] = [], shown = text, meta: PromptMeta = {}): Promise<"stop" | "steps" | "budget" | "cancelled" | "error"> {
@@ -618,13 +661,13 @@ export class Engine {
     this.busy = true
     this.o.bus.emit({ type: "prompt", text: shown, ...(meta.turnId ? { turnId: meta.turnId } : {}), ...(meta.attachments?.length ? { attachments: meta.attachments } : {}), ...(meta.machine ? { machine: true as const } : {}) })
     try {
-      return await this.run(text, [...this.pendingContext.splice(0), ...extra], meta.turnId, meta.machine)
+      return await this.run(text, [...this.pendingContext.splice(0), ...extra], meta.turnId, meta.machine, meta.maxSteps)
     } finally {
       this.busy = false
     }
   }
 
-  private async run(text: string, extra: (string | ImagePart)[], turnId?: string, machine?: true): Promise<"stop" | "steps" | "budget" | "cancelled" | "error"> {
+  private async run(text: string, extra: (string | ImagePart)[], turnId?: string, machine?: true, stepCap?: number): Promise<"stop" | "steps" | "budget" | "cancelled" | "error"> {
     const { bus } = this.o
     this.abort = new AbortController()
     const signal = this.abort.signal
@@ -648,7 +691,7 @@ export class Engine {
     this.turnSteers = []
     this.autoCompactions = 0
     const toolMap = new Map(this.o.tools.map((t) => [t.name, t]))
-    const maxSteps = this.o.maxSteps ?? LOOP.steps.default
+    const maxSteps = Math.min(this.o.maxSteps ?? LOOP.steps.default, stepCap ?? Infinity)
     const started = performance.now()
     this.turnWaitedMs = 0
     this.turnOutputBytes = 0
@@ -666,7 +709,10 @@ export class Engine {
         if (ranOut) bus.emit({ type: "notice", message: `stopped ${ranOut.said} — the model answers from what it has; send a message to carry on` })
       }
       const wrapUp = step > maxSteps || ranOut !== undefined
-      if (!wrapUp) await this.manageContext(signal)
+      if (!wrapUp && (await this.manageContext(signal)) === "stop") {
+        bus.emit({ type: "done", reason: "stop" })
+        return "stop"
+      }
       // After compacting, so what was just sent is never folded into a summary.
       if (!wrapUp) this.deliverInbox()
       bus.emit({ type: "step", n: step })
@@ -809,11 +855,11 @@ export class Engine {
         ...this.o.toolCtx,
         signal,
         backgroundNow: bgNow.signal,
-        question: this.o.asker.question ? (req, id) => this.waiting(this.o.asker.question!(req, id)) : undefined,
+        question: this.o.asker.question ? (req, id) => this.waiting(this.o.asker.question!(req, id), "question") : undefined,
         plan: {
           get: () => this.plan,
           set: (p) => this.setPlan(p, "model"),
-          ask: this.o.asker.plan ? (req, id) => this.waiting(this.o.asker.plan!(req, id)) : undefined,
+          ask: this.o.asker.plan ? (req, id) => this.waiting(this.o.asker.plan!(req, id), "plan") : undefined,
         },
         setMode: (m) => {
           this.mode = m
@@ -873,6 +919,11 @@ export class Engine {
       // No dialect takes an image inside a tool result the same way; a user turn right after the
       // results works everywhere.
       if (images.length) this.push({ role: "user", parts: [{ type: "text", text: "The image(s) you asked to view:" }, ...images], turn: false })
+      // Spec 3.4.0: a subagent that called team_finish has ended — nothing runs after that call.
+      if (this.o.team?.ended?.()) {
+        bus.emit({ type: "done", reason: "stop" })
+        return "stop"
+      }
       if (signal.aborted) {
         bus.emit({ type: "done", reason: "cancelled" })
         return "cancelled"
@@ -943,6 +994,13 @@ export class Engine {
     let edited: string | undefined
     const ctx: PermContext = { ...this.o.perm, mode: this.mode, rules: [...this.o.perm.rules, ...this.sessionRules] }
     let decision = evaluate(request, ctx)
+    // Spec 3.4.0 (subagents.md §7): a subagent of the tree — itself, the main agent, or the person.
+    if (team?.permit) {
+      const { mode: _m, ...rest } = ctx
+      const v = team.permit(request, rest)
+      if (v.answer === "refused") return { result: { output: v.text, isError: true }, denied: true }
+      decision = { ...v.decision, action: v.answer === "run" ? "allow" : "ask" }
+    }
 
     // Doom-loop guard: the same call three times running is asked about, whatever the mode.
     const key = `${call.name} ${JSON.stringify(args)}`
@@ -959,7 +1017,7 @@ export class Engine {
           (diff) => ({ diff }),
           (e) => ({ error: (e as Error).message }),
         )
-      const reply = await this.waiting(this.o.asker.ask({ tool: tool.name, args, request, decision, preview, purpose }))
+      const reply = await this.waiting(this.o.asker.ask({ tool: tool.name, args, request, decision, preview, purpose }), "approval", tool.name)
       if (reply.kind === "deny" && looping)
         return {
           result: {

src/session/progress.ts

@@ -0,0 +1,240 @@
+// Progress (harness spec 3.4.0, harness/progress.md): one fixed set of phases every client shows
+// the same way. The object, its text from loop.json `progress.texts` and the one line drawn for it.
+import { LOOP } from "../harness.ts"
+import { fillPrompt } from "../prompt/fill.ts"
+import { duration } from "../duration.ts"
+
+const P = LOOP.progress
+
+export type Phase =
+  | "queued"
+  | "waiting_for_model"
+  | "loading_model"
+  | "reading_prompt"
+  | "thinking"
+  | "writing"
+  | "running_tool"
+  | "waiting_for_person"
+  | "waiting_for_subagents"
+  | "compacting"
+  | "retrying"
+  | "stopped"
+  | "failed"
+
+/** The structured progress object: the phase, when it began (Unix ms), and its data. */
+export type Progress = { phase: Phase; since: number } & {
+  ahead?: number
+  model?: string
+  reason?: string
+  done?: number
+  total?: number
+  rate?: number
+  tool?: string
+  what?: string
+  running?: number
+  in?: number
+  attempt?: number
+}
+
+export const PHASES = P.phases as readonly Phase[]
+
+const fill = (t: string, v: Record<string, string | number>) => fillPrompt(t, Object.fromEntries(Object.entries(v).map(([k, x]) => [k, String(x)])))
+
+/** A call's short title as `what`: one line, at most what_max characters. */
+export function shortWhat(s: string | undefined): string | undefined {
+  if (!s) return undefined
+  const line = s.split("\n")[0]!.trim()
+  if (!line) return undefined
+  return line.length > P.what_max ? `${line.slice(0, P.what_max - 1)}…` : line
+}
+
+/** The phase's text (progress.md §2), its fields filled; `now` for retrying's countdown. */
+export function progressText(p: Progress, now = Date.now()): string {
+  const T = P.texts as Record<string, string>
+  switch (p.phase) {
+    case "queued":
+      return p.ahead !== undefined ? fill(T["queued.ahead"]!, { ahead: p.ahead }) : T.queued!
+    case "waiting_for_model":
+      return fill(T.waiting_for_model!, { model: p.model ?? "", reason: (P.reason as Record<string, string>)[p.reason ?? ""] ?? p.reason ?? "" })
+    case "loading_model":
+      return fill(T.loading_model!, { model: p.model ?? "" })
+    case "reading_prompt":
+      if (p.total === undefined || p.done === undefined || p.total <= 0) return T["reading_prompt.bare"]!
+      return fill(T.reading_prompt!, { done: p.done, total: p.total, percent: Math.floor((p.done * 100) / p.total) })
+    case "writing":
+      return p.rate !== undefined && Number.isFinite(p.rate) ? fill(T.writing!, { rate: p.rate.toFixed(1) }) : T["writing.bare"]!
+    case "running_tool":
+      return p.what ? fill(T.running_tool!, { tool: p.tool ?? "", what: p.what }) : fill(T["running_tool.bare"]!, { tool: p.tool ?? "" })
+    case "waiting_for_person": {
+      const W = P.what as Record<string, string>
+      const what = p.what === "approval" ? (p.tool ? fill(W.approval!, { tool: p.tool }) : W["approval.bare"]!) : (W[p.what ?? ""] ?? p.what ?? "")
+      return fill(T.waiting_for_person!, { what })
+    }
+    case "waiting_for_subagents":
+      return fill(T.waiting_for_subagents!, { running: p.running ?? 0 })
+    case "retrying": {
+      const left = Math.max(0, (p.in ?? 0) - Math.floor((now - p.since) / 1000))
+      return fill(T.retrying!, { in: left, reason: p.reason ?? "" })
+    }
+    case "stopped":
+      return p.reason ? fill(T["stopped.reason"]!, { reason: (P.stopped_reason as Record<string, string>)[p.reason] ?? p.reason }) : T.stopped!
+    case "failed":
+      return fill(T.failed!, { reason: p.reason ?? "" })
+    default:
+      return T[p.phase] ?? p.phase
+  }
+}
+
+/** The line a client draws (progress.md §3): the text and the time since it began, or the text
+ *  alone for stopped and failed. */
+export function progressLine(p: Progress, now = Date.now()): string {
+  const text = progressText(p, now)
+  if ((P.no_elapsed as string[]).includes(p.phase)) return text
+  return fill(P.line, { text, elapsed: duration(now - p.since, true) })
+}
+
+/** Whether two objects are the same phase for the same thing (so `since` is kept). */
+export function samePhase(a: Progress | null | undefined, b: Omit<Progress, "since">): boolean {
+  if (!a || a.phase !== b.phase) return false
+  if (b.phase === "running_tool") return a.tool === b.tool && a.what === b.what
+  if (b.phase === "waiting_for_model" || b.phase === "loading_model") return a.model === b.model && a.reason === b.reason
+  if (b.phase === "waiting_for_person") return a.what === b.what && a.tool === b.tool
+  if (b.phase === "retrying") return a.attempt === b.attempt && a.reason === b.reason
+  return true
+}
+
+/** What a tool call does, in a line (`what` of running_tool): the command, the path, the query. */
+export function callWhat(args: unknown): string | undefined {
+  if (!args || typeof args !== "object") return undefined
+  const a = args as Record<string, unknown>
+  for (const k of ["command", "path", "file_path", "pattern", "query", "url", "description", "name"]) if (typeof a[k] === "string" && a[k]) return shortWhat(a[k] as string)
+  return undefined
+}
+
+type BusLike = { on(fn: (e: any) => void): () => void; emit(e: any): void }
+
+/** The progress of one engine, derived from the events it emits (progress.md §4), and said on its
+ *  bus as `phase` events: a new phase at once, a change of its data at most pace_per_second times a
+ *  second. Explicit phases (compacting, waiting for the person or a model, queued) are `set` by
+ *  whoever knows them. */
+export class ProgressTracker {
+  current: Progress | null = null
+  private rate: number | undefined
+  private running = new Map<string, Omit<Progress, "since">>()
+  private lastEmit = 0
+  private pendingTimer: ReturnType<typeof setTimeout> | undefined
+  private error: string | undefined
+  private held: Progress | null = null
+  private off: () => void
+
+  constructor(
+    private bus: BusLike,
+    private now: () => number = Date.now,
+  ) {
+    this.off = bus.on((e) => this.on(e))
+  }
+
+  dispose() {
+    this.off()
+    if (this.pendingTimer) clearTimeout(this.pendingTimer)
+  }
+
+  /** Set the phase (or null). Same phase for the same thing keeps `since`. */
+  set(next: Omit<Progress, "since"> | null) {
+    if (next === null) {
+      if (this.current === null) return
+      this.current = null
+      return this.emit(true)
+    }
+    if (samePhase(this.current, next)) {
+      const merged = { ...this.current!, ...next, since: this.current!.since } as Progress
+      if (JSON.stringify(merged) === JSON.stringify(this.current)) return
+      this.current = merged
+      return this.emit(false)
+    }
+    this.current = { ...next, since: this.now() } as Progress
+    this.emit(true)
+  }
+
+  /** A phase that waits on something outside the turn (the person, a model): set now, and the
+   *  phase before it put back by the function returned. */
+  hold(next: Omit<Progress, "since">): () => void {
+    const before = this.current
+    this.held = before
+    this.set(next)
+    return () => {
+      if (this.current && this.current.phase === next.phase) {
+        this.current = before ? { ...before } : null
+        this.emit(true)
+      }
+      this.held = null
+    }
+  }
+
+  private emit(now: boolean) {
+    const t = this.now()
+    const gap = 1000 / LOOP.progress.pace_per_second
+    if (!now && t - this.lastEmit < gap) {
+      this.pendingTimer ??= setTimeout(() => {
+        this.pendingTimer = undefined
+        this.emit(true)
+      }, gap - (t - this.lastEmit))
+      return
+    }
+    if (this.pendingTimer) {
+      clearTimeout(this.pendingTimer)
+      this.pendingTimer = undefined
+    }
+    this.lastEmit = t
+    this.bus.emit({ type: "phase", progress: this.current ? { ...this.current } : null })
+  }
+
+  private on(e: any) {
+    switch (e.type) {
+      case "task":
+        if (e.state === "start") {
+          this.rate = undefined
+          this.running.clear()
+          this.error = undefined
+        }
+        return
+      case "step":
+        this.running.clear()
+        return this.set({ phase: "reading_prompt" })
+      case "progress":
+        return this.set({ phase: "reading_prompt", done: (e.cache ?? 0) + (e.processed ?? 0), total: e.total })
+      case "reasoning":
+        return this.set({ phase: "thinking" })
+      case "text":
+      case "tool_call_delta":
+        return this.set(this.rate !== undefined ? { phase: "writing", rate: this.rate } : { phase: "writing" })
+      case "usage":
+        if (!e.child && typeof e.rate === "number" && Number.isFinite(e.rate)) this.rate = e.rate
+        return
+      case "tool_start": {
+        const a = (e.args ?? {}) as Record<string, unknown>
+        const p: Omit<Progress, "since"> = e.name === "task" && a.background !== true ? { phase: "waiting_for_subagents", running: 1 } : { phase: "running_tool", tool: e.name, ...(callWhat(e.args) ? { what: callWhat(e.args) } : {}) }
+        this.running.set(e.id, p)
+        if (this.running.size === 1) this.set(p)
+        return
+      }
+      case "tool_end": {
+        this.running.delete(e.id)
+        const next = this.running.values().next()
+        if (!next.done) this.set(next.value)
+        return
+      }
+      case "notice":
+        if (e.retry) this.set({ phase: "retrying", reason: shortWhat(e.retry.reason) ?? "", in: e.retry.in, ...(e.retry.attempt ? { attempt: e.retry.attempt } : {}) })
+        return
+      case "error":
+        this.error = shortWhat(String(e.message ?? "")) ?? "error"
+        return
+      case "done":
+        this.running.clear()
+        if (e.reason === "stop") return this.set(null)
+        if (e.reason === "error") return this.set({ phase: "failed", reason: this.error ?? "error" })
+        return this.set({ phase: "stopped", reason: e.reason === "cancelled" ? "person" : e.reason })
+    }
+  }
+}

src/team/delegation.ts

@@ -0,0 +1,86 @@
+// Delegation (harness spec 3.4.0, subagents.md §7): a subagent's mode is its parent's or stricter,
+// request options go with a stricter mode, and who decides a subagent's call — itself, the main
+// agent (through a request), or the person. Pure rules over permission/evaluate.ts.
+import { LOOP } from "../harness.ts"
+import { fillPrompt } from "../prompt/fill.ts"
+import type { Mode } from "../config/schema.ts"
+import { evaluate, type Context as PermContext, type Decision, type PermissionRequest, type Rule, type ToolClass } from "../permission/evaluate.ts"
+import MODES_JSON from "../../harness/permission/modes.json"
+
+const TEXTS = LOOP.team.texts
+const fill = (t: string, v: Record<string, string>) => fillPrompt(t, v)
+
+/** Loosest first (permission/modes.json `order`). */
+export const MODE_ORDER = MODES_JSON.order as Mode[]
+const rank = (m: Mode) => MODE_ORDER.indexOf(m)
+
+/** a is looser than b. */
+export const looser = (a: Mode, b: Mode) => rank(a) < rank(b)
+/** a is stricter than b. */
+export const stricter = (a: Mode, b: Mode) => rank(a) > rank(b)
+/** The stricter of the two. */
+export const strictest = (a: Mode, b: Mode): Mode => (stricter(a, b) ? a : b)
+
+export const REQUEST_KINDS = LOOP.team.request.kinds as ToolClass[]
+
+/** team_add's / team_assign's mode and request options for a parent in `parentMode` (§7): the
+ *  refusal, or the mode and options the subagent gets. */
+export function checkMode(parentMode: Mode, mode: Mode | undefined, request: string[] | undefined): { error: string } | { mode: Mode; request: ToolClass[] } {
+  const m = mode ?? parentMode
+  if (looser(m, parentMode)) return { error: fill(TEXTS.wider, { parent_mode: parentMode }) }
+  const req = request ?? []
+  if (req.length && !stricter(m, parentMode)) return { error: TEXTS.request_not_narrower }
+  const bad = req.find((k) => !(REQUEST_KINDS as string[]).includes(k))
+  if (bad !== undefined) return { error: fill(TEXTS.request_kind, { kind: bad }) }
+  return { mode: m, request: [...new Set(req)] as ToolClass[] }
+}
+
+/** A grant the main agent gave (team_decide): a rule for this subagent alone. */
+export interface Grant {
+  tool: string
+  /** The permission key of the tool (edit and write share `edit`). */
+  permission?: string
+  pattern: string
+}
+
+export type Verdict = { answer: "run"; decision: Decision } | { answer: "person"; decision: Decision } | { answer: "refused"; text: string; decision?: Decision }
+
+/** Who decides a subagent's call (§7). `ctx` is the session's permission context (its rules and
+ *  approvals); `mode` the subagent's, `main` the main agent's. A grant applies only where the main
+ *  agent's own mode would allow the call. */
+export function whoDecides(
+  req: PermissionRequest,
+  o: { ctx: Omit<PermContext, "mode">; mode: Mode; main: Mode; request: readonly string[]; grants: readonly Grant[]; unattended?: boolean },
+): Verdict {
+  const mainD = evaluate(req, { ...o.ctx, mode: o.main })
+  const grantRules: Rule[] = mainD.action === "allow" ? o.grants.map((g) => ({ permission: g.permission ?? g.tool, pattern: g.pattern, action: "allow" as const })) : []
+  const own = evaluate(req, { ...o.ctx, mode: o.mode, rules: [...o.ctx.rules, ...grantRules] })
+  if (own.action === "allow") return { answer: "run", decision: own }
+  if (own.action === "deny") return { answer: "refused", text: `Denied: ${own.reason}`, decision: own }
+  if (mainD.action === "allow") return { answer: "refused", text: o.request.includes(req.class) ? TEXTS.requestable : TEXTS.not_delegated }
+  if (mainD.action === "deny") return { answer: "refused", text: `Denied: ${mainD.reason}`, decision: mainD }
+  if (o.unattended) return { answer: "refused", text: "Refused: nobody can approve this in an unattended session.", decision: own }
+  return { answer: "person", decision: own }
+}
+
+/** team_request's own check: the tool's risk class must be one of the options. */
+export function requestRefusal(tool: string, cls: ToolClass | undefined, request: readonly string[]): string | undefined {
+  if (cls && request.includes(cls)) return undefined
+  return fill(TEXTS.not_requestable, { tool, kinds: request.join(", ") || "none" })
+}
+
+/** The main agent made stricter: every subagent looser than its parent is made as strict as it,
+ *  down the tree (§7). `agents` in tree order; returns each one's mode after. */
+export function tighten(main: Mode, agents: { name: string; parent: string; mode: Mode }[], lead = LOOP.team.names.lead): Record<string, Mode> {
+  const modes: Record<string, Mode> = { [lead]: main }
+  const out: Record<string, Mode> = {}
+  const pending = [...agents]
+  // Parents first: an agent's parent is the lead or one listed before it, but tolerate any order.
+  for (let guard = 0; pending.length && guard < 1000; guard++) {
+    const i = pending.findIndex((a) => modes[a.parent] !== undefined)
+    if (i < 0) break
+    const a = pending.splice(i, 1)[0]!
+    modes[a.name] = out[a.name] = strictest(a.mode, modes[a.parent]!)
+  }
+  return out
+}

src/team/runtime.ts

@@ -1,5 +1,5 @@
-// A lead session's team at work (harness spec 3.2.0, team.md): the rules of team.ts driven by real
+// A lead session's team at work (harness spec 3.2.0, team.md; 3.4.0, subagents.md §5–§13): the
-// turns. The lead is the session's own engine; each member is a subagent engine of its own (the job
+// rules of team.ts driven by real turns. The lead is the session's own engine; each member is a subagent engine of its own (the job
 // machinery, app.ts spawn), started when the team gives it a slot. What the rules decide happens
 // through three effects: a turn starts (a member's brief, a woken agent's messages), a message reaches
 // a working agent at its next step, an agent is stopped.
@@ -8,10 +8,14 @@ import { TOOL_SPECS } from "../harness.ts"
 import type { Engine } from "../session/engine.ts"
 import type { Tool } from "../tool/tool.ts"
 import { parallelFrom, Team, TEAM, type TeamAgent } from "./team.ts"
+import type { Mode } from "../config/schema.ts"
+import type { ToolClass } from "../permission/evaluate.ts"
 
 /** A member's engine, as the app made it. */
 export interface MemberHandle {
   engine: Engine
+  /** Its one final turn after its context filled (spec 3.4.0 §9), with this text, few steps. */
+  final?(text: string): void
   /** Its next turn, with this text as the prompt (its context intact). */
   run(text: string): void
   /** Stop it where it is; nothing more comes from it. */
@@ -25,10 +29,12 @@ export interface RuntimeOptions {
   /** Make a member's engine (attached here with `attach`) and start its first turn with `text`. */
   spawnMember(member: TeamAgent, text: string): void
   /** The lead: a message at its next step (`note`), or — idle — a turn of its own (`wake`). */
-  lead: { note(text: string): void; wake(text: string): void }
+  lead: { note(text: string): void; wake(text: string): void; /** Spec 3.4.0: its progress while it waits for a slot. */ waiting?(model: string): void }
   /** An unknown agent or model, answered as task answers it; undefined when both are known. The
    *  model's ref as the session knows it. */
   resolve?(agent: string, model: string | undefined): { error: string } | { model: string }
+  /** A tool's risk class and permission key, by its name (team_request, team_decide). */
+  tools?: () => readonly Tool[]
 }
 
 export class TeamRuntime {
@@ -36,6 +42,11 @@ export class TeamRuntime {
   private members = new Map<string, MemberHandle>()
   /** The lead's turn waiting for a slot: resolved when it is given one. */
   private leadWaiter: (() => void) | undefined
+  /** Spec 3.4.0: subagents that ended by their own team_finish (their turn stops after the call),
+   *  and the final turn waiting for one whose context filled. */
+  private ended = new Set<string>()
+  private finals = new Map<string, string>()
+  private pace: ReturnType<typeof setTimeout> | undefined
 
   constructor(private o: RuntimeOptions) {
     this.team = o.team
@@ -47,6 +58,8 @@ export class TeamRuntime {
       start: (name, text) => this.start(name, text),
       deliver: (name, text) => this.deliver(name, text),
       stop: (name) => this.stopped(name),
+      end: (name) => void this.ended.add(name),
+      final: (name, text) => void this.finals.set(name, text),
     }
   }
 
@@ -59,8 +72,82 @@ export class TeamRuntime {
     this.changed()
   }
 
+  /** Said on the lead's bus as `team`, at most four times a second (acp.md `team` event). */
   private changed() {
+    // A change inside the pace window is said once at its end, so the last state always goes out.
+    if (this.pace) return void (this.trailing = true)
     this.o.bus.emit({ type: "team" })
+    this.pace = setTimeout(() => {
+      this.pace = undefined
+      if (this.trailing) {
+        this.trailing = false
+        this.changed()
+      }
+    }, 1000 / 4)
+  }
+  private trailing = false
+
+  /** Something changed that a reader must see now (the tools' answers, tests): said at once. */
+  private changedNow() {
+    if (this.pace) clearTimeout(this.pace)
+    this.pace = undefined
+    this.trailing = false
+    this.o.bus.emit({ type: "team" })
+  }
+
+  /** A tool's risk class, for team_request. */
+  classOf(tool: string): ToolClass | undefined {
+    const spec = TOOL_SPECS[tool]?.risk
+    if (spec) return spec as ToolClass
+    const t = this.o.tools?.().find((x) => x.name === tool)
+    try {
+      return t ? (t.permission({} as never, {} as never).class as ToolClass) : undefined
+    } catch {
+      return "execute"
+    }
+  }
+
+  /** A tool's permission key (edit and write share `edit`), for a grant. */
+  permissionKeyOf(tool: string): string | undefined {
+    return this.o.tools?.().find((x) => x.name === tool)?.access ?? tool
+  }
+
+  /** Every subagent's engine in its team's mode (spec 3.4.0 §7: the main agent made stricter, an
+   *  assignment). */
+  syncModes() {
+    for (const [name, h] of this.members) {
+      const m = this.team.get(name)?.mode
+      if (m && h.engine.mode !== m) h.engine.mode = m
+    }
+  }
+
+  /** The main agent's mode changed (the person, plan approved): stricter reaches down the tree. */
+  mainMode(mode: Mode) {
+    this.team.setMainMode(mode)
+    this.syncModes()
+    this.changed()
+  }
+
+  /** Spec 3.4.0: whether this subagent ended by its own call (its engine stops after the step). */
+  hasEnded(name: string): boolean {
+    return this.ended.has(name)
+  }
+
+  /** Spec 3.4.0 §9: its context reached pause_at. "stop" the first time (it is paused, its final
+   *  turn follows), "go" in that final turn (no compaction, it carries on). */
+  contextFull(name: string): "stop" | "go" {
+    const a = this.team.get(name)
+    if (!a || a.state === "paused") return "go"
+    this.team.pause(name)
+    this.changedNow()
+    return "stop"
+  }
+
+  /** What a member's run should do when its turn ended: its final turn (context full), or report. */
+  takeFinal(name: string): string | undefined {
+    const t = this.finals.get(name)
+    this.finals.delete(name)
+    return t
   }
 
   /** Something the tools did changed the team: said, for whoever shows it. */
@@ -70,7 +157,7 @@ export class TeamRuntime {
 
   /** team_add for `as`: an unknown agent or model is answered as task answers it, after the
    *  team's own refusals (the name, the goal, the size), which come first. */
-  add(as: string, a: { name: string; role: string; task: string; goal?: string; lead?: boolean; agent?: string; model?: string }): { text: string; added?: TeamAgent } {
+  add(as: string, a: { name: string; role: string; task: string; goal?: string; lead?: boolean; agent?: string; model?: string; parent?: string; mode?: Mode; request?: string[] }): { text: string; added?: TeamAgent } {
     const agent = a.agent ?? "general"
     const r = this.o.resolve?.(agent, a.model) ?? { model: a.model ?? this.team.get(as)?.model ?? this.team.model }
     // The team's refusals first; an added member needs a known agent and model.
@@ -111,6 +198,7 @@ export class TeamRuntime {
   private stopped(name: string) {
     const h = this.members.get(name)
     this.members.delete(name)
+    this.finals.delete(name)
     h?.cancel()
     this.changed()
   }
@@ -125,6 +213,7 @@ export class TeamRuntime {
       return true
     }
     this.o.bus.emit({ type: "notice", message: `waiting for a free slot: the model serves ${parallelFrom(this.team.capacityOf(this.team.model))} requests at once and the team's members hold them` })
+    this.o.lead.waiting?.(this.team.model)
     this.changed()
     return await new Promise<boolean>((resolve) => {
       const onAbort = () => {
@@ -147,8 +236,21 @@ export class TeamRuntime {
     this.changed()
   }
 
-  /** A member's turn ended: its reply is its report; messages it was given too late wake it again. */
+  /** A member's turn ended: its reply is its report; messages it was given too late wake it again.
+   *  Spec 3.4.0: one that ended by its own team_finish is let go; one whose context filled gets its
+   *  final turn first. */
   memberDone(name: string, reply: string, reason: string, leftover: string[]) {
+    if (this.ended.delete(name)) {
+      this.members.delete(name)
+      this.changed()
+      return
+    }
+    const final = this.takeFinal(name)
+    if (final !== undefined && this.team.get(name)?.state === "paused") {
+      this.members.get(name)?.final?.(final)
+      this.changed()
+      return
+    }
     if (!this.team.get(name)) return
     this.team.finish(name, reply, reason)
     if (leftover.length) this.team.undelivered(name, leftover)
@@ -157,7 +259,6 @@ export class TeamRuntime {
 
   /** What a call answers before its permission check, for the agent `as`. */
   refuse(as: string, tool: Tool, args: unknown): string | undefined {
-    if (tool.name === "task" && as !== TEAM.names.lead) return undefined
     // A tool the spec does not have (an MCP server's) by the class its call carries: read when it
     // says readOnlyHint, interact for the server's memory, else execute.
     let risk: string | undefined

src/team/subagents.ts

@@ -0,0 +1,101 @@
+// Subagents (harness spec 3.3.0, harness/subagents.md): which model a subagent — a task subagent or
+// a team's member — runs on, and whether subagents are offered at all. The rules alone, without an
+// engine: the conformance area `subagents` drives them, and the app (spawn, team_add) asks them.
+import { LOOP } from "../harness.ts"
+import type { Message } from "../provider/types.ts"
+import { parallelFrom, type Capacity } from "./team.ts"
+
+export const SUBAGENTS = LOOP.subagents
+const TEXTS = SUBAGENTS.texts
+
+/** A model the person may use: its served id (the ref), its id at its connection, its display name. */
+export interface ModelInfo {
+  ref: string
+  id?: string
+  name?: string
+  capacity: Capacity
+}
+
+const escape = (s: string) => s.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")
+const BEFORE = new RegExp(`${SUBAGENTS.named.before}$`, "u")
+const AFTER = new RegExp(`^(?:${SUBAGENTS.named.after})`, "u")
+
+/** The names a model goes by: the served id, the part after its provider, the display name. */
+export function namesOf(m: Pick<ModelInfo, "ref" | "id" | "name">): string[] {
+  const out = [m.ref]
+  if (m.id) out.push(m.id)
+  const slash = m.ref.lastIndexOf("/")
+  if (slash >= 0) out.push(m.ref.slice(slash + 1))
+  if (m.name?.trim()) out.push(m.name.trim())
+  return [...new Set(out.filter(Boolean))]
+}
+
+/** Whether `name` occurs in `text` as a whole word, without case (subagents.md §2). White space in
+ *  the name matches any run of white space. */
+export function occursAsWord(text: string, name: string): boolean {
+  const pattern = name
+    .trim()
+    .split(/\s+/)
+    .map(escape)
+    .join("\\s+")
+  if (!pattern) return false
+  const re = new RegExp(pattern, "giu")
+  for (let m = re.exec(text); m; m = re.exec(text)) {
+    const before = text.slice(0, m.index)
+    const after = text.slice(m.index + m[0].length)
+    if (!BEFORE.test(before) && !AFTER.test(after)) return true
+    re.lastIndex = m.index + 1
+  }
+  return false
+}
+
+/** Whether the person named the model in any of the messages they typed. */
+export function named(m: Pick<ModelInfo, "ref" | "id" | "name">, said: readonly string[]): boolean {
+  const names = namesOf(m)
+  return said.some((t) => names.some((n) => occursAsWord(t, n)))
+}
+
+/** What the person typed in a session, oldest first: their prompts and what they sent mid-turn
+ *  (after `steerMark`) — never a note, a job's or a team's marked message, or a machine turn. */
+export function personTexts(messages: readonly Message[], steerMark: string): string[] {
+  const out: string[] = []
+  for (const m of messages) {
+    if (m.role !== "user" || m.machine) continue
+    const first = m.parts[0]
+    if (!first || first.type !== "text") continue
+    if (m.turn === false) {
+      if (first.text.startsWith(steerMark)) out.push(first.text.slice(steerMark.length).trim())
+      continue
+    }
+    out.push(first.text)
+  }
+  return out
+}
+
+/** Whether a model can be its own session's subagent: it serves two requests at once, or has no limit. */
+export function servesOwn(c: Capacity): boolean {
+  return parallelFrom(c) >= 2
+}
+
+/** subagents.md §4: whether subagents are offered for a session on `session`, given the models the
+ *  person may use, and whether the session's model can serve its own. */
+export function offer(session: ModelInfo, models: readonly ModelInfo[], subagents = true): { offered: boolean; own: boolean } {
+  const own = servesOwn(session.capacity)
+  const group = session.capacity.group
+  const other = models.some((m) => m.ref !== session.ref && (group === undefined || m.capacity.group !== group))
+  return { offered: subagents && (own || other), own }
+}
+
+/** subagents.md §2, §3: the model a subagent runs on — the session's, or `wanted` (already resolved
+ *  to a model the person may use) — or the refusal, word for word. */
+export function subagentModel(session: ModelInfo, wanted: ModelInfo | undefined, said: readonly string[]): { model: string } | { refused: string } {
+  const m = wanted ?? session
+  if (m.ref !== session.ref) {
+    if (!named(m, said)) return { refused: TEXTS.not_named }
+    const group = session.capacity.group
+    if (group !== undefined && m.capacity.group === group) return { refused: TEXTS.same_group }
+    return { model: m.ref }
+  }
+  if (!servesOwn(session.capacity)) return { refused: TEXTS.one_request }
+  return { model: session.ref }
+}

src/team/team.ts

@@ -1,5 +1,7 @@
-// Teams (harness spec 3.2.0, harness/team.md): a lead session's agents, their room, who receives
+// Teams (harness spec 3.2.0, harness/team.md; 3.4.0, harness/subagents.md §5–§13): the main agent
-// what, the size the person allows, a manager's refusals, and the slots of the model they run on.
+// (`lead`) and its subagents, their room, who receives what, the size the person allows, a
+// manager's refusals, the slots of the model they run on — and from 3.4.0 the tree only the main
+// agent builds, modes and request options, findings, and a subagent's end.
 // This file is the rules alone, without an engine: the conformance area `team` drives it step by
 // step, and the app (team/runtime.ts) drives it from real turns — the effects say when an agent's
 // turn starts (it was given a slot), when a message reaches an agent that is working, and when an
@@ -9,14 +11,42 @@ import { fillPrompt } from "../prompt/fill.ts"
 import { DEFAULT_RULES, evaluate, toRules } from "../permission/evaluate.ts"
 import { BUILTIN_HARDLINE } from "../permission/hardline.ts"
 import { duration } from "../duration.ts"
+import type { Mode } from "../config/schema.ts"
+import type { ToolClass } from "../permission/evaluate.ts"
+import { checkMode, strictest, type Grant } from "./delegation.ts"
+import type { Progress } from "../session/progress.ts"
 
 export const TEAM = LOOP.team
 const TEXTS = TEAM.texts
 const LEAD = TEAM.names.lead
-export const TEAM_TOOLS = ["team_add", "team_read", "team_send", "team_stop"] as const
+export const TEAM_TOOLS = ["team_add", "team_assign", "team_decide", "team_findings", "team_finish", "team_read", "team_replace", "team_request", "team_send", "team_stop", "team_whoami"] as const
+/** Spec 3.4.0 (subagents.md §8): the main agent's alone. */
+export const MAIN_ONLY = TEAM.main_only.tools as readonly string[]
+/** The main agent's team tools, and a subagent's (with team_request only when it has request options). */
+export const LEAD_TOOLS = ["team_add", "team_assign", "team_decide", "team_findings", "team_read", "team_replace", "team_send", "team_stop", "team_whoami"] as const
+export const MEMBER_TOOLS = ["team_findings", "team_finish", "team_read", "team_send", "team_whoami"] as const
+export function memberTools(a: { request: readonly string[] }): string[] {
+  return [...MEMBER_TOOLS, ...(a.request.length ? ["team_request"] : [])].sort()
+}
 const NAME = new RegExp(TEAM.names.pattern)
 
-export type AgentState = "working" | "waiting" | "finished" | "idle"
+export type AgentState = "working" | "waiting" | "finished" | "idle" | "paused" | "done" | "out_of_context" | "stopped"
+export type EndState = "done" | "out_of_context" | "stopped"
+
+export interface Finding {
+  n: number
+  text: string
+  time: number
+}
+
+export interface Request {
+  id: string
+  from: string
+  tool: string
+  pattern: string
+  why: string
+  open: boolean
+}
 
 /** What a model's server can do at once (models/schema.json capacity). */
 export interface Capacity {
@@ -32,11 +62,26 @@ export function parallelFrom(c: Capacity | undefined): number {
 }
 
 export interface TeamAgent {
+  /** `lead` for the main agent; s1, s2, … for subagents, never reused (spec 3.4.0). */
+  id: string
   name: string
+  /** Its position (team_add's role). */
   role: string
+  /** Its task, as given and as assigned since. */
+  task: string
+  mode: Mode
+  /** Request options (spec 3.4.0 §7). */
+  request: ToolClass[]
+  /** What the main agent allowed it (team_decide). */
+  grants: Grant[]
+  findings: Finding[]
+  /** Its context use, as its engine last said it. */
+  context?: { used: number; window: number }
+  /** Its current progress (progress.md). */
+  progress?: Progress | null
   /** Who added it; null for the lead. */
   manager: string | null
-  /** May add members of its own (team_add's `lead`); the lead may. */
+  /** A sub-manager (team_add's `lead`), or the main agent: subagents may be placed under it. */
   leads: boolean
   /** The team's goal as it was told it (the lead's: the last one it gave). */
   goal: string
@@ -54,6 +99,13 @@ export interface TeamAgent {
   ended?: number
   /** Added in this order (tree order among siblings). */
   seq: number
+  /** How it ended, and when (an ended subagent stays in the tree's `ended`). */
+  endState?: EndState
+  endedAt?: number
+  /** The sub-manager it was under before that one ended (team_replace puts it back). */
+  movedFrom?: string
+  /** The device's job that runs it (its transcript, read-only). */
+  jobId?: string
 }
 
 export interface RoomMessage {
@@ -65,6 +117,38 @@ export interface RoomMessage {
   text: string
   time: number
   report: boolean
+  /** Spec 3.4.0: what it is — a message, a report, an assignment, a request, a decision. */
+  kind: RoomKind
+}
+
+export type RoomKind = "message" | "report" | "assignment" | "request" | "decision"
+
+/** A row of the tree (subagents.md §12). */
+export interface TreeAgent {
+  id: string
+  name: string
+  position: string
+  parent: string | null
+  state: AgentState
+  model: string
+  mode: Mode
+  request: string[]
+  manager: boolean
+  task?: string
+  context?: { used: number; window: number }
+  progress: Progress | null
+  findings: number
+  messages: number
+  ended?: number
+}
+
+export interface TreeInfo {
+  min: number
+  max: number
+  size: number
+  open_slots: number
+  agents: TreeAgent[]
+  ended: TreeAgent[]
 }
 
 export interface TeamEffects {
@@ -75,6 +159,11 @@ export interface TeamEffects {
   deliver(name: string, text: string): void
   /** It left the team: its turn stops where it is, its background commands with it. */
   stop(name: string): void
+  /** Spec 3.4.0: it ended by its own team_finish — its turn stops after that call. */
+  end?(name: string): void
+  /** Spec 3.4.0: its context is full — its turn stops after this step, and its one final turn
+   *  starts with `text`. */
+  final?(name: string, text: string): void
 }
 
 export interface TeamOptions {
@@ -86,12 +175,19 @@ export interface TeamOptions {
   /** The lead's model (the chat's), and every model's capacity. */
   model: string
   capacity: (model: string) => Capacity
+  /** Spec 3.3.0 (subagents.md §4): every model the person may use, by ref — a model off the lead's
+   *  one-model server can serve as a subagent where the lead's model cannot. None: only the lead's. */
+  models?: () => readonly string[]
   effects?: TeamEffects
   /** A bash command a manager may still run: one the default rules alone allow in manual mode. */
   readingBash?: (command: string) => boolean
   now?: () => number
   /** The lead's state as the team begins: working (its first prompt runs) or idle. */
   leadState?: "working" | "idle"
+  /** The main agent's mode (the session's). Default: auto. */
+  mode?: Mode
+  /** Steps a turn may take: the main agent's and a subagent's (for team_whoami). */
+  steps?: { main: number; sub: number }
 }
 
 /** The default rules' answer for a command in manual mode, with no rules of the person's. */
@@ -111,10 +207,10 @@ export class Team {
   model: string
   private agents = new Map<string, TeamAgent>()
   room: RoomMessage[] = []
-  /** `task` helpers running in the team: who started each, and on which model. */
+  /** `task` subagents running in the team: who started each, and on which model. */
-  private helpers: { id: number; by: string; model: string }[] = []
+  private tasks: { id: number; by: string; model: string }[] = []
   private seq = 0
-  private helperSeq = 0
+  private taskSeq = 0
   /** Whether the team is offered, fixed as the prompt is made (§2). */
   offered = false
   /** The setting as the lead was last told it. */
@@ -122,6 +218,11 @@ export class Team {
   /** Members the person stopped, not yet said to the lead. */
   private stoppedByPerson: string[] = []
   private waits = 0
+  /** Subagents that ended (stopped, done, out of context), in the order they ended. */
+  ended: TeamAgent[] = []
+  private idSeq = 0
+  private reqSeq = 0
+  requests: Request[] = []
 
   constructor(private o: TeamOptions) {
     this.cap = o.cap ?? TEAM.agents.cap
@@ -130,7 +231,7 @@ export class Team {
     this.min = TEAM.agents.min
     this.max = TEAM.agents.max
     this.setSetting(o.min ?? TEAM.agents.min, o.max ?? TEAM.agents.max)
-    this.agents.set(LEAD, { name: LEAD, role: TEAM.read.lead_role, manager: null, leads: true, goal: "", model: o.model, agent: "", state: o.leadState ?? "working", inbox: [], seq: this.seq++ })
+    this.agents.set(LEAD, { id: LEAD, name: LEAD, role: TEAM.read.lead_role, task: "", mode: o.mode ?? "auto", request: [], grants: [], findings: [], manager: null, leads: true, goal: "", model: o.model, agent: "", state: o.leadState ?? "working", inbox: [], seq: this.seq++ })
     this.fix()
   }
 
@@ -152,7 +253,7 @@ export class Team {
 
   /** Where it is set: a value outside 1 ≤ min ≤ max ≤ cap is refused with the bounds. */
   static check(min: number, max: number, cap: number): string | undefined {
-    if (!Number.isInteger(min) || !Number.isInteger(max) || min < 1 || min > max || max > cap) return `agents: 1 ≤ min ≤ max ≤ ${cap} (the cap) — whole numbers; min ${min}, max ${max} is outside`
+    if (!Number.isInteger(min) || !Number.isInteger(max) || min < 1 || min > max || max > cap) return `Subagents: 1 ≤ Min ≤ Max ≤ ${cap} (the cap), whole numbers, this agent included; Min ${min}, Max ${max} is outside`
     return undefined
   }
 
@@ -165,13 +266,22 @@ export class Team {
     return parallelFrom(this.o.capacity(this.model))
   }
 
-  /** §2: offered when subagents are on and the chat's model serves two at once — fixed as the
+  /** Spec 3.3.0 (subagents.md §4): whether a model could serve as a subagent now — the lead's own,
-   *  prompt is made (and made again on a model switch); `offered` is a kept prompt's decision. */
+   *  with two slots or more, or another model off its one-model server — and whether it is the own. */
+  couldServe(): { any: boolean; own: boolean } {
+    const own = this.parallelNow() >= 2
+    const group = this.o.capacity(this.model).group
+    const other = (this.o.models?.() ?? []).some((r) => r !== this.model && (group === undefined || this.o.capacity(r).group !== group))
+    return { any: own || other, own }
+  }
+
+  /** §2: offered when subagents are on and a model could serve as a subagent — fixed as the prompt
+   *  is made (and made again on a model switch); `offered` is a kept prompt's decision. */
   fix(model = this.model, offered?: boolean): boolean {
     this.model = model
     const lead = this.agents.get(LEAD)
     if (lead) lead.model = model
-    this.offered = offered ?? (this.subagents && this.parallelNow() >= 2)
+    this.offered = offered ?? (this.subagents && this.couldServe().any)
     return this.offered
   }
 
@@ -207,9 +317,9 @@ export class Team {
     return this.agents.size > 1
   }
 
-  /** §3: the lead, every member, and every running `task` helper. */
+  /** §3: the lead, every member, and every running `task` subagent. */
   size(): number {
-    return this.agents.size + this.helpers.length
+    return this.agents.size + this.tasks.length
   }
 
   /** §6: an agent with a member, or the lead from the start when min ≥ 2 and the team is offered. */
@@ -222,8 +332,9 @@ export class Team {
 
   /** §2: the team tools' requirement `team`; the text a call answers while it is not met. */
   unavailable(): string | undefined {
-    if ((this.max >= 2 && this.parallelNow() >= 2) || this.hasMembers()) return undefined
+    const could = this.couldServe().any
-    return this.parallelNow() < 2 ? TEAM.unavailable.one_request : TEAM.unavailable.one_agent
+    if ((this.max >= 2 && could) || this.hasMembers()) return undefined
+    return !could ? TEAM.unavailable.one_request : TEAM.unavailable.one_agent
   }
 
   /** §3: what `task` answers in a team, or undefined when it may run (and then counts). */
@@ -250,8 +361,14 @@ export class Team {
   /** What a call by `by` answers before it runs — the team tools' requirement, task's size, a
    *  manager's work — or undefined when it goes on to its own checks (permission first). */
   refusal(by: string, tool: string, args: unknown, risk?: string): string | undefined {
-    if ((TEAM_TOOLS as readonly string[]).includes(tool)) return this.offered ? this.unavailable() : TEAM.unavailable.one_agent
+    if ((TEAM_TOOLS as readonly string[]).includes(tool)) {
-    if (tool === "task") return this.taskRefusal()
+      const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
+      if (no) return no
+      // Spec 3.4.0 (subagents.md §6): only the main agent adds, assigns, replaces, stops, decides.
+      if (by !== LEAD && MAIN_ONLY.includes(tool)) return TEXTS.only_main
+      return undefined
+    }
+    if (tool === "task") return by !== LEAD ? TEXTS.only_main : this.taskRefusal()
     if (this.isManager(by) && this.isWork(tool, args, risk)) return TEXTS.manager
     return undefined
   }
@@ -259,14 +376,14 @@ export class Team {
   // ── slots ───────────────────────────────────────────────────────────────────
 
   private busyOn(model: string): number {
-    return [...this.agents.values()].filter((a) => a.state === "working" && a.model === model).length + this.helpers.filter((h) => h.model === model).length
+    return [...this.agents.values()].filter((a) => a.state === "working" && a.model === model).length + this.tasks.filter((h) => h.model === model).length
   }
 
   private free(model: string): boolean {
     const group = this.o.capacity(model).group
     if (group !== undefined) {
       for (const a of this.agents.values()) if (a.state === "working" && a.model !== model && this.o.capacity(a.model).group === group) return false
-      for (const h of this.helpers) if (h.model !== model && this.o.capacity(h.model).group === group) return false
+      for (const h of this.tasks) if (h.model !== model && this.o.capacity(h.model).group === group) return false
     }
     return this.busyOn(model) < parallelFrom(this.o.capacity(model))
   }
@@ -299,22 +416,22 @@ export class Team {
     return lead.state === "working"
   }
 
-  /** A `task` helper started by `by` (it counts and holds a slot until helperDone). */
+  /** A `task` subagent started by `by` (it counts and holds a slot until taskDone). */
-  helperStart(by: string, model = this.model): number {
+  taskStart(by: string, model = this.model): number {
-    const id = ++this.helperSeq
+    const id = ++this.taskSeq
-    this.helpers.push({ id, by, model })
+    this.tasks.push({ id, by, model })
     return id
   }
 
-  helperDone(id: number) {
+  taskDone(id: number) {
-    this.helpers = this.helpers.filter((h) => h.id !== id)
+    this.tasks = this.tasks.filter((h) => h.id !== id)
     this.fillSlots()
   }
 
-  /** The first helper `by` started, done (the conformance area's `helper_done`). */
+  /** The first task subagent `by` started, done (the conformance area's `task_done`). */
-  helperDoneBy(by: string) {
+  taskDoneBy(by: string) {
-    const h = this.helpers.find((x) => x.by === by)
+    const h = this.tasks.find((x) => x.by === by)
-    if (h) this.helperDone(h.id)
+    if (h) this.taskDone(h.id)
   }
 
   // ── delivery ────────────────────────────────────────────────────────────────
@@ -323,6 +440,8 @@ export class Team {
   private deliver(name: string, text: string) {
     const a = this.agents.get(name)
     if (!a) return
+    // Spec 3.4.0 §9: a paused subagent is given nothing new; the message stays in the room.
+    if (a.state === "paused") return
     this.delivered.push({ name, text })
     if (a.state === "working") return this.o.effects?.deliver(name, text)
     a.inbox.push(text)
@@ -332,10 +451,11 @@ export class Team {
   /** What was delivered, to whom, since this was last read (for a front end and the conformance). */
   delivered: { name: string; text: string }[] = []
 
-  private post(from: string, to: string[], recipients: string[], text: string, report: boolean): RoomMessage {
+  private post(from: string, to: string[], recipients: string[], text: string, report: boolean, kind: RoomKind = report ? "report" : "message", id = ""): RoomMessage {
-    const m: RoomMessage = { n: this.room.length + 1, from, to, recipients, text, time: this.now(), report }
+    const m: RoomMessage = { n: this.room.length + 1, from, to, recipients, text, time: this.now(), report, kind }
     this.room.push(m)
-    const mark = fill(report ? TEAM.marks.report : TEAM.marks.message, { from, to: to.join(", ") })
+    const marks = TEAM.marks as Record<string, string>
+    const mark = fill(marks[kind]!, { from, to: to.join(", "), id })
     for (const r of recipients) this.deliver(r, `${mark}\n${text}`)
     return m
   }
@@ -343,32 +463,49 @@ export class Team {
   // ── the tools ───────────────────────────────────────────────────────────────
 
   /** team_add's refusals, in the spec's order (§4), or undefined when it would add. */
-  wouldRefuse(by: string, a: { name: string; goal?: string }): string | undefined {
+  wouldRefuse(by: string, a: { name: string; goal?: string; parent?: string; mode?: Mode; request?: string[] }): string | undefined {
     const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
     if (no) return no
+    if (by !== LEAD) return TEXTS.only_main
     if (a.name === "all") return TEXTS.all_name
     if (!NAME.test(a.name)) return fill(TEXTS.bad_name, { name: a.name })
-    if (this.agents.has(a.name)) return fill(TEXTS.taken, { name: a.name })
+    if (this.agents.has(a.name) || a.name === TEAM.names.parent) return fill(TEXTS.taken, { name: a.name })
-    if (by === LEAD && !this.agents.get(LEAD)!.goal && !a.goal?.trim()) return TEXTS.no_goal
+    if (!this.agents.get(LEAD)!.goal && !a.goal?.trim()) return TEXTS.no_goal
     if (this.size() + 1 > this.max) return fill(TEXTS.full, { max: this.max })
-    return undefined
+    return this.placeRefusal(a.parent, a.mode, a.request)
+  }
+
+  /** Spec 3.4.0 (§6, §7): `parent` must be the lead or a sub-manager; the mode no looser than its. */
+  private placeRefusal(parent: string | undefined, mode: Mode | undefined, request: string[] | undefined): string | undefined {
+    const p = this.agents.get(parent ?? LEAD)
+    if (!p) return fill(TEXTS.unknown, { name: parent!, names: this.names() })
+    if (!p.leads) return fill(TEXTS.not_manager, { name: p.name })
+    const m = checkMode(p.mode, mode, request)
+    return "error" in m ? m.error : undefined
   }
 
   /** team_add (§4). `brief` is what the member's first turn is given. */
-  add(by: string, a: { name: string; role: string; task: string; goal?: string; lead?: boolean; agent?: string; model?: string }): { text: string; added?: TeamAgent } {
+  add(by: string, a: { name: string; role: string; task: string; goal?: string; lead?: boolean; agent?: string; model?: string; parent?: string; mode?: Mode; request?: string[] }): { text: string; added?: TeamAgent } {
     const no = this.wouldRefuse(by, a)
     if (no) return { text: no }
     const lead = this.agents.get(LEAD)!
     const goal = a.goal?.trim()
-    const caller = this.agents.get(by) ?? lead
+    const parent = this.agents.get(a.parent ?? LEAD)!
-    if (by === LEAD && goal) lead.goal = goal
+    if (goal) lead.goal = goal
+    const placed = checkMode(parent.mode, a.mode, a.request) as { mode: Mode; request: ToolClass[] }
     const member: TeamAgent = {
+      id: `${TEAM.names.id_prefix}${++this.idSeq}`,
       name: a.name,
       role: a.role.trim().split("\n")[0]!,
-      manager: by,
+      task: a.task,
+      mode: placed.mode,
+      request: placed.request,
+      grants: [],
+      findings: [],
+      manager: parent.name,
       leads: a.lead === true,
-      goal: goal || caller.goal,
+      goal: goal || parent.goal || lead.goal,
-      model: a.model ?? caller.model,
+      model: a.model ?? lead.model,
       agent: a.agent ?? "general",
       state: "finished",
       inbox: [],
@@ -386,7 +523,13 @@ export class Team {
   send(by: string, a: { to: string[] | string; message: string }): string {
     const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
     if (no) return no
-    const list = [...new Set(typeof a.to === "string" ? [a.to] : a.to)]
+    let list = [...new Set(typeof a.to === "string" ? [a.to] : a.to)]
+    // Spec 3.4.0: `parent` is the sender's parent.
+    if (list.includes(TEAM.names.parent)) {
+      const mgr = this.agents.get(by)?.manager
+      if (!mgr) return TEXTS.no_parent
+      list = [...new Set(list.map((n) => (n === TEAM.names.parent ? mgr : n)))]
+    }
     const all = list.includes("all")
     const others = [...this.agents.keys()].filter((n) => n !== by)
     if (all) {
@@ -439,12 +582,21 @@ export class Team {
         fill(R.agent, {
           indent: "  ".repeat(depth(a)),
           name: a.name,
+          id: a.id,
           role: a.role,
           state: R.states[a.state],
           model: a.model,
+          mode: a.request.length ? `${a.mode} ${fill(R.request, { kinds: a.request.join(", ") })}` : a.mode,
+          context: a.context && a.context.window > 0 ? fill(R.context, { percent: Math.floor((a.context.used * 100) / a.context.window) }) : "—",
           last: a.state === "working" ? R.active : a.ended !== undefined ? `${duration(this.now() - a.ended)} ago` : "—",
         }),
       )
+    const open = this.openSlots()
+    if (open) lines.push(fill(R.slots, { n: open }))
+    if (this.ended.length) {
+      lines.push(R.ended)
+      for (const a of this.ended) lines.push(fill(R.ended_agent, { name: a.name, id: a.id, role: a.role, state: R.states[a.state], findings: a.findings.length }))
+    }
     const { messages, more } = this.readRoom(by, since)
     lines.push("")
     if (!messages.length) lines.push(fill(R.none, { since }))
@@ -464,6 +616,7 @@ export class Team {
   stopCmd(by: string, name: string): string {
     const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
     if (no) return no
+    if (by !== LEAD) return TEXTS.only_main
     if (name === "all") {
       const under = this.membersOf(by)
       if (!under.length) return TEXTS.nobody_to_stop
@@ -487,10 +640,26 @@ export class Team {
     return gone.map((a) => a.name)
   }
 
-  private remove(gone: TeamAgent[]) {
+  private remove(gone: TeamAgent[], end: EndState = "stopped") {
     for (const a of gone) {
       this.agents.delete(a.name)
-      this.o.effects?.stop(a.name)
+      a.state = end
+      a.endState = end
+      a.endedAt = this.now()
+      a.inbox = []
+      a.brief = undefined
+      this.ended.push(a)
+      for (const r of this.requests) if (r.from === a.name) r.open = false
+      if (end === "done") this.o.effects?.end?.(a.name)
+      else if (end === "stopped") this.o.effects?.stop(a.name)
+      // Spec 3.4.0: the subagents under one that ended by itself move under its parent until the
+      // main agent replaces it (team_replace moves them back under the replacement).
+      if (end !== "stopped")
+        for (const c of this.agents.values())
+          if (c.manager === a.name) {
+            c.manager = a.manager ?? LEAD
+            c.movedFrom = a.name
+          }
     }
     this.fillSlots()
   }
@@ -499,7 +668,16 @@ export class Team {
    *  stop, steps, budget, or `error — <text>`. */
   finish(name: string, reply = "", reason = "stop") {
     const a = this.agents.get(name)
-    if (!a || a.state !== "working") return
+    if (!a) return
+    // Spec 3.4.0 §9: the final turn after its context filled has ended — it is out of context.
+    if (a.state === "paused") {
+      const gone = fill(TEXTS.gone, { name: a.name, id: a.id })
+      const text = reply.trim() ? `${gone}\n\n${reply.trim()}` : gone
+      if (a.manager) this.post(name, [a.manager], [a.manager], text, true)
+      this.remove([a], "out_of_context")
+      return
+    }
+    if (a.state !== "working") return
     a.state = name === LEAD ? "idle" : "finished"
     a.ended = this.now()
     if (a.manager) {
@@ -534,6 +712,241 @@ export class Team {
     this.stoppedByPerson = []
   }
 
+
+  // ── spec 3.4.0: the main agent's tree (subagents.md §5–§13) ─────────────────────
+
+  /** A subagent by name (live first, else the latest that ended) or by id. */
+  find(nameOrId: string): TeamAgent | undefined {
+    return this.agents.get(nameOrId) ?? [...this.agents.values()].find((a) => a.id === nameOrId) ?? [...this.ended].reverse().find((a) => a.name === nameOrId || a.id === nameOrId)
+  }
+
+  /** Open slots (§6): min − 1 places for subagents not added yet, while min ≥ 2 and offered. */
+  openSlots(): number {
+    if (!this.offered || this.min < 2) return 0
+    return Math.max(0, this.min - 1 - (this.agents.size - 1))
+  }
+
+  /** The main agent's mode changed: stricter reaches down the tree, looser changes nobody (§7). */
+  setMainMode(mode: Mode) {
+    this.agents.get(LEAD)!.mode = mode
+    for (const a of this.tree()) if (a.manager) a.mode = strictest(a.mode, this.agents.get(a.manager)?.mode ?? mode)
+  }
+
+  /** Its engine's context use, and its progress, as last said. */
+  setContext(name: string, used: number, window: number) {
+    const a = this.agents.get(name)
+    if (a) a.context = { used, window }
+  }
+
+  setProgress(name: string, p: Progress | null) {
+    const a = this.agents.get(name)
+    if (a) a.progress = p
+  }
+
+  /** team_whoami (§8). */
+  whoami(by: string): string {
+    const a = this.agents.get(by)
+    if (!a) return fill(TEXTS.unknown, { name: by, names: this.names() })
+    const W = TEAM.whoami.lines
+    const steps = by === LEAD ? (this.o.steps?.main ?? LOOP.steps.default) : (this.o.steps?.sub ?? LOOP.jobs.subagent_steps.default)
+    const lines = [fill(W.who, { name: a.name, id: a.id, role: a.role })]
+    if (a.manager) lines.push(fill(W.parent, { parent: a.manager }))
+    if (a.manager) lines.push(fill(W.task, { task: a.task.split("\n")[0]! }))
+    lines.push(fill(W.mode, { mode: a.mode }))
+    if (a.request.length) lines.push(fill(W.request, { kinds: a.request.join(", ") }))
+    lines.push(fill(W.model, { model: a.model }))
+    lines.push(fill(W.limits, { size: this.size(), max: this.max, steps, spawn: by === LEAD ? W.spawn_yes : W.spawn_no }))
+    if (a.context && a.context.window > 0) lines.push(fill(W.context, { used: a.context.used, window: a.context.window, percent: Math.floor((a.context.used * 100) / a.context.window) }))
+    return lines.join("\n")
+  }
+
+  /** team_findings write (§8). */
+  noteFinding(by: string, text: string): string {
+    const a = this.agents.get(by)
+    if (!a) return fill(TEXTS.unknown, { name: by, names: this.names() })
+    const n = a.findings.length + 1
+    a.findings.push({ n, text: text.trim(), time: this.now() })
+    return fill(TEXTS.noted, { n })
+  }
+
+  /** Whether `by` may read `target`'s findings: its own, one in its subtree, or the lead anyone's. */
+  private mayRead(by: string, target: TeamAgent): boolean {
+    if (by === LEAD || target.name === by) return true
+    let up: string | null | undefined = target.endState ? (target.movedFrom ?? target.manager) : target.manager
+    for (let guard = 0; up && guard < 1000; guard++) {
+      if (up === by) return true
+      up = this.agents.get(up)?.manager ?? this.find(up)?.manager
+    }
+    return false
+  }
+
+  /** team_findings read (§8). */
+  readFindings(by: string, of?: string): string {
+    const target = of ? this.find(of) : this.agents.get(by)
+    if (!target) return fill(TEXTS.unknown, { name: of ?? by, names: this.names() })
+    if (!this.mayRead(by, target)) return TEXTS.not_yours_findings
+    const lines = [fill(TEAM.findings.header, { name: target.name, id: target.id })]
+    if (!target.findings.length) lines.push(TEXTS.no_findings)
+    for (const f of target.findings) lines.push(fill(TEAM.findings.item, { n: f.n, text: f.text }))
+    return lines.join("\n")
+  }
+
+  /** team_finish (§8): the report to its parent, and it ends (done). */
+  finishTask(by: string, report: string): string {
+    const a = this.agents.get(by)
+    if (!a || !a.manager) return TEXTS.no_parent
+    this.post(by, [a.manager], [a.manager], report.trim() || TEXTS.no_report, true)
+    const to = a.manager
+    this.remove([a], "done")
+    return fill(TEXTS.finished, { to })
+  }
+
+  /** Its context is full (§9): paused, and given one final turn. The text of that turn. */
+  pause(name: string): string | undefined {
+    const a = this.agents.get(name)
+    if (!a || a.name === LEAD || a.state === "paused") return undefined
+    a.state = "paused"
+    a.inbox = []
+    const text = `${fill(TEAM.marks.context_full, { to: name })}\n${fill(TEXTS.context_full, { steps: TEAM.context.final_steps })}`
+    this.delivered.push({ name, text })
+    this.o.effects?.final?.(name, text)
+    return text
+  }
+
+  /** team_assign (§8). */
+  assign(by: string, a: { name: string; task: string; role?: string; parent?: string; mode?: Mode }): string {
+    const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
+    if (no) return no
+    if (by !== LEAD) return TEXTS.only_main
+    const t = this.agents.get(a.name)
+    if (!t || t.name === LEAD) return this.find(a.name)?.endState ? fill(TEXTS.has_ended, { name: a.name }) : fill(TEXTS.unknown, { name: a.name, names: this.names() })
+    const parentName = a.parent ?? t.manager!
+    const p = this.agents.get(parentName)
+    if (!p) return fill(TEXTS.unknown, { name: parentName, names: this.names() })
+    if (!p.leads) return fill(TEXTS.not_manager, { name: p.name })
+    if (this.subtree(t.name).some((x) => x.name === p.name)) return fill(TEXTS.cycle, { name: t.name, parent: p.name })
+    let mode = strictest(t.mode, p.mode)
+    let request = t.request
+    if (a.mode) {
+      const m = checkMode(p.mode, a.mode, t.request.length && a.mode !== p.mode ? t.request : undefined)
+      if ("error" in m) return m.error
+      mode = m.mode
+      request = m.mode === p.mode ? [] : t.request
+    }
+    t.task = a.task
+    if (a.role?.trim()) t.role = a.role.trim().split("\n")[0]!
+    t.manager = p.name
+    t.movedFrom = undefined
+    t.mode = mode
+    t.request = request
+    for (const c of this.subtree(t.name).slice(1)) c.mode = strictest(c.mode, this.agents.get(c.manager!)!.mode)
+    this.post(LEAD, [t.name], [t.name], a.task, false, "assignment")
+    this.fillSlots()
+    return fill(TEXTS.assigned, { name: t.name })
+  }
+
+  /** team_replace (§8): the same name, position, parent, mode and model; a new id; briefed with
+   *  the task and the old one's findings. */
+  replace(by: string, a: { name: string; task?: string }): { text: string; added?: TeamAgent } {
+    const no = this.offered ? this.unavailable() : TEAM.unavailable.one_agent
+    if (no) return { text: no }
+    if (by !== LEAD) return { text: TEXTS.only_main }
+    const old = this.agents.get(a.name) ?? [...this.ended].reverse().find((x) => x.name === a.name)
+    if (!old || old.name === LEAD) return { text: fill(TEXTS.unknown, { name: a.name, names: this.names() }) }
+    const live = this.agents.get(a.name) === old
+    if (!live && this.size() + 1 > this.max) return { text: fill(TEXTS.full, { max: this.max }) }
+    if (live) {
+      // It alone ends: the subagents under it move to the replacement.
+      for (const c of this.membersOf(old.name)) c.movedFrom = old.name
+      for (const c of this.membersOf(old.name)) c.manager = old.manager ?? LEAD
+      this.remove([old], "stopped")
+    }
+    const manager = old.movedFrom ?? old.manager
+    const parent = manager && this.agents.get(manager) ? manager : LEAD
+    const findings = old.findings.length ? old.findings.map((f) => fill(TEAM.findings.item, { n: f.n, text: f.text })).join("\n") : TEXTS.no_findings
+    const brief = fill(TEXTS.replace_brief, { task: a.task?.trim() || old.task, name: old.name, id: old.id, findings })
+    const p = this.agents.get(parent)!
+    const fresh: TeamAgent = {
+      ...old,
+      id: `${TEAM.names.id_prefix}${++this.idSeq}`,
+      task: a.task?.trim() || old.task,
+      mode: strictest(old.mode, p.mode),
+      grants: [],
+      findings: [],
+      context: undefined,
+      progress: undefined,
+      manager: parent,
+      state: "finished",
+      inbox: [],
+      brief,
+      seq: this.seq++,
+      endState: undefined,
+      endedAt: undefined,
+      movedFrom: undefined,
+      ended: undefined,
+      waitSince: undefined,
+      jobId: undefined,
+    }
+    this.agents.set(fresh.name, fresh)
+    for (const c of this.agents.values())
+      if (c.movedFrom === old.name) {
+        c.manager = fresh.name
+        c.movedFrom = undefined
+      }
+    this.want(fresh)
+    this.fillSlots()
+    return { text: fill(TEXTS.replaced, { name: fresh.name, id: fresh.id, size: this.size(), max: this.max }), added: fresh }
+  }
+
+  /** team_request (§7): `cls` is the tool's risk class. */
+  request(by: string, a: { tool: string; pattern: string; why: string }, cls: ToolClass | undefined): string {
+    const t = this.agents.get(by)
+    if (!t || !t.manager) return TEXTS.no_parent
+    if (!cls || !t.request.includes(cls)) return fill(TEXTS.not_requestable, { tool: a.tool, kinds: t.request.join(", ") || "none" })
+    const id = `${TEAM.request.id_prefix}${++this.reqSeq}`
+    this.requests.push({ id, from: by, tool: a.tool, pattern: a.pattern, why: a.why, open: true })
+    this.post(by, [LEAD], [LEAD], fill(TEXTS.request_body, { tool: a.tool, pattern: a.pattern, why: a.why }), false, "request", id)
+    this.fillSlots()
+    return fill(TEXTS.requested, { id })
+  }
+
+  /** team_decide (§7): a grant for the subagent when allowed; it is told either way. `permission`:
+   *  the tool's permission key. */
+  decide(by: string, a: { id: string; allow: boolean; note?: string }, permission?: (tool: string) => string | undefined): string {
+    if (by !== LEAD) return TEXTS.only_main
+    const r = this.requests.find((x) => x.id === a.id && x.open)
+    const t = r && this.agents.get(r.from)
+    if (!r || !t) return fill(TEXTS.no_request, { id: a.id })
+    r.open = false
+    if (a.allow) t.grants.push({ tool: r.tool, permission: permission?.(r.tool) ?? r.tool, pattern: r.pattern })
+    const head = fill(a.allow ? TEXTS.decision_allowed : TEXTS.decision_refused, { tool: r.tool, pattern: r.pattern })
+    this.post(LEAD, [t.name], [t.name], a.note?.trim() ? `${head}\n\n${a.note.trim()}` : head, false, "decision", r.id)
+    this.fillSlots()
+    return fill(TEXTS.decided, { decision: a.allow ? TEXTS.allowed : TEXTS.refused, id: r.id, name: t.name })
+  }
+
+  /** The tree as clients show it (§12). */
+  info(): TreeInfo {
+    const row = (a: TeamAgent): TreeAgent => ({
+      id: a.id,
+      name: a.name,
+      position: a.role,
+      parent: a.manager,
+      state: a.state,
+      model: a.model,
+      mode: a.mode,
+      request: [...a.request],
+      manager: a.name === LEAD ? true : a.leads,
+      ...(a.manager ? { task: a.task } : {}),
+      ...(a.context ? { context: { ...a.context } } : {}),
+      progress: a.progress ?? null,
+      findings: a.findings.length,
+      messages: this.room.filter((m) => m.from === a.name || m.recipients.includes(a.name)).length,
+      ...(a.endedAt !== undefined ? { ended: a.endedAt } : {}),
+    })
+    return { min: this.min, max: this.max, size: this.size(), open_slots: this.openSlots(), agents: this.tree().map(row), ended: this.ended.map(row) }
+  }
+
   // ── what the lead is told ───────────────────────────────────────────────────
 
   /** §10: the note's entries for the lead's next user message — the size when it was not said yet
@@ -545,7 +958,7 @@ export class Team {
     if (!this.told || this.told.min !== this.min || this.told.max !== this.max) {
       this.told = { min: this.min, max: this.max }
       if (this.max === 1) out.push(N.team_one)
-      else out.push([fill(N.team, { min: this.min, max: this.max }), ...(this.min >= 2 ? [fill(N.team_lead, { members: this.min - 1 })] : [])].join("\n"))
+      else out.push([fill(N.team, { min: this.min, max: this.max }), ...(this.min >= 2 ? [fill(N.team_lead, { members: this.min - 1 })] : []), ...(this.couldServe().own ? [] : [N.team_own])].join("\n"))
     }
     if (this.stoppedByPerson.length) out.push(fill(N.team_stopped, { names: this.stoppedByPerson.splice(0).join(", ") }))
     return out

src/tool/registry.ts

@@ -11,7 +11,7 @@ import { planSubmitTool } from "./plan_submit.ts"
 import { planUpdateTool } from "./plan_update.ts"
 import { planWriteTool } from "./plan_write.ts"
 import { taskTool } from "./task.ts"
-import { teamAddTool, teamReadTool, teamSendTool, teamStopTool } from "./team_tools.ts"
+import { teamAddTool, teamAssignTool, teamDecideTool, teamFindingsTool, teamFinishTool, teamReadTool, teamReplaceTool, teamRequestTool, teamSendTool, teamStopTool, teamWhoamiTool } from "./team_tools.ts"
 import { decisionsTool, tasksTool } from "./project_tools.ts"
 import { webFetchTool, webSearchTool } from "./web.ts"
 import { memoryTool } from "./memory.ts"
@@ -46,6 +46,14 @@ export const BUILTIN_TOOLS: Tool[] = [
   teamSendTool,
   teamReadTool,
   teamStopTool,
+  // Spec 3.4.0 (subagents.md §8).
+  teamWhoamiTool,
+  teamFindingsTool,
+  teamFinishTool,
+  teamRequestTool,
+  teamDecideTool,
+  teamAssignTool,
+  teamReplaceTool,
   planWriteTool,
   planUpdateTool,
   tasksTool,

src/tool/team_tools.ts

@@ -1,5 +1,6 @@
-// The team tools (harness spec 3.2.0, team.md; tools/team_*.json): add a member, send through the
+// The team tools (harness spec 3.2.0, team.md; 3.4.0, subagents.md §5–§13; tools/team_*.json):
-// room, read the team and the room, stop a member. The rules are team/team.ts; these are its calls
+// add a subagent, send through the room, read the tree and the room, stop one — and from 3.4.0
+// whoami, findings, finish, request and decide, assign and replace. The rules are team/team.ts; these are its calls
 // for the agent the context names (`ctx.team.as`).
 import { z } from "zod"
 import { defineTool, ToolError, type ToolContext } from "./tool.ts"
@@ -37,6 +38,9 @@ export const teamAddTool = defineTool({
     lead: z.boolean().optional(),
     agent: z.string().optional(),
     model: z.string().optional(),
+    parent: z.string().optional(),
+    mode: z.enum(["auto", "edit", "manual", "plan"]).optional(),
+    request: z.array(z.enum(["read", "write", "execute"])).optional(),
   }),
   invalid: badName,
   permission: (a) => ({ permission: "team_add", class: "interact", patterns: [a.agent ?? "general"] }),
@@ -92,3 +96,113 @@ export const teamStopTool = defineTool({
     return { output, title: t.runtime.team.size() < size ? `stopped ${a.name}` : "not stopped", isError: t.runtime.team.size() >= size }
   },
 })
+
+export const teamWhoamiTool = defineTool({
+  ...described("team_whoami"),
+  purpose: false,
+  access: "team_whoami",
+  requires: "team",
+  schema: z.object({}),
+  permission: () => ({ permission: "team_whoami", class: "read", patterns: ["*"] }),
+  async run(_a, ctx) {
+    const t = teamOf(ctx)
+    return { output: t.runtime.team.whoami(t.as), title: t.as }
+  },
+})
+
+export const teamFindingsTool = defineTool({
+  ...described("team_findings"),
+  purpose: false,
+  access: "team_findings",
+  requires: "team",
+  schema: z.object({ action: z.enum(["write", "read"]), text: z.string().optional(), of: z.string().optional() }),
+  permission: (a) => ({ permission: "team_findings", class: "interact", patterns: [a.action] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    if (a.action === "write") {
+      if (!a.text?.trim()) throw new ToolError("write needs `text`: the finding.")
+      const output = t.runtime.team.noteFinding(t.as, a.text)
+      t.runtime.notify()
+      return { output, title: "finding noted" }
+    }
+    return { output: t.runtime.team.readFindings(t.as, a.of), title: `findings of ${a.of ?? t.as}` }
+  },
+})
+
+export const teamFinishTool = defineTool({
+  ...described("team_finish"),
+  purpose: false,
+  access: "team_finish",
+  requires: "team",
+  schema: z.object({ report: z.string().min(1) }),
+  permission: () => ({ permission: "team_finish", class: "interact", patterns: ["*"] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    const output = t.runtime.team.finishTask(t.as, a.report)
+    t.runtime.notify()
+    return { output, title: "finished" }
+  },
+})
+
+export const teamRequestTool = defineTool({
+  ...described("team_request"),
+  purpose: false,
+  access: "team_request",
+  requires: "team",
+  schema: z.object({ tool: z.string().min(1), pattern: z.string().min(1), why: z.string().min(1) }),
+  permission: () => ({ permission: "team_request", class: "interact", patterns: ["*"] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    const before = t.runtime.team.requests.length
+    const output = t.runtime.team.request(t.as, a, t.runtime.classOf(a.tool))
+    t.runtime.notify()
+    const asked = t.runtime.team.requests.length > before
+    return { output, title: asked ? `asked for ${a.tool} ${a.pattern}` : "not asked", isError: !asked }
+  },
+})
+
+export const teamDecideTool = defineTool({
+  ...described("team_decide"),
+  purpose: false,
+  access: "team_decide",
+  requires: "team",
+  schema: z.object({ id: z.string().min(1), allow: z.boolean(), note: z.string().optional() }),
+  permission: () => ({ permission: "team_decide", class: "interact", patterns: ["*"] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    const output = t.runtime.team.decide(t.as, a, (tool) => t.runtime.permissionKeyOf(tool))
+    t.runtime.notify()
+    return { output, title: `${a.allow ? "allowed" : "refused"} ${a.id}` }
+  },
+})
+
+export const teamAssignTool = defineTool({
+  ...described("team_assign"),
+  purpose: false,
+  access: "team_assign",
+  requires: "team",
+  schema: z.object({ name: z.string().min(1), task: z.string().min(1), role: z.string().optional(), parent: z.string().optional(), mode: z.enum(["auto", "edit", "manual", "plan"]).optional() }),
+  permission: (a) => ({ permission: "team_assign", class: "interact", patterns: [a.name] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    const output = t.runtime.team.assign(t.as, a)
+    t.runtime.syncModes()
+    t.runtime.notify()
+    return { output, title: `to ${a.name}` }
+  },
+})
+
+export const teamReplaceTool = defineTool({
+  ...described("team_replace"),
+  purpose: false,
+  access: "team_replace",
+  requires: "team",
+  schema: z.object({ name: z.string().min(1), task: z.string().optional() }),
+  permission: (a) => ({ permission: "team_replace", class: "interact", patterns: [a.name] }),
+  async run(a, ctx) {
+    const t = teamOf(ctx)
+    const r = t.runtime.team.replace(t.as, a)
+    t.runtime.notify()
+    return { output: r.text, title: r.added ? `${a.name} → ${r.added.id}` : "not replaced", isError: !r.added }
+  },
+})

src/tui/app.tsx

@@ -129,6 +129,8 @@ export function Root(p: { app: Runtime; view: View; theme: Theme; onQuit: () =>
   const sendToJob = (text: string) => {
     const id = entered()!
     const job = app.registry.get(app.jobsSessionId(), id)
+    // Spec 3.4.0 §11: a subagent of the tree is read-only; the person speaks through the main agent.
+    if (job?.team) return view.info(`${id} is ${job.team}, a subagent of the main agent: its session is read-only — tell the main agent, or /team stop ${job.team}`, "warn")
     if (!job?.subagent) return view.info(`${id} cannot be messaged any more — an earlier process ran it`, "warn")
     if (job.state === "running") {
       job.subagent.send(text)

src/tui/commands.ts

@@ -11,11 +11,13 @@ import { doRelease, planRelease } from "../git/release.ts"
 import { instructionFiles } from "../prompt/assemble.ts"
 import { addDecision, readBoard, readDecisions, renderBoard } from "../project/board.ts"
 import { contextBreakdown, renderBreakdown } from "../session/context.ts"
+import { SUBAGENTS } from "../team/subagents.ts"
+import { TEAM } from "../team/team.ts"
 import { changelogPrompt, initPrompt, reviewPrompt, teamText } from "../session/commands.ts"
 import { git } from "../git/run.ts"
 import { branches, deleteBranch, switchBranch } from "../git/branch.ts"
 import type { Ctx } from "./context.ts"
-import type { DialogSpec } from "./components/dialog.tsx"
+import type { DialogSpec, Option } from "./components/dialog.tsx"
 import { deleteEverywhere } from "../acp/share.ts"
 import { themeNames } from "./theme.ts"
 import { render } from "../plan/plan.ts"
@@ -85,7 +87,6 @@ const pendingCodes = new Map<string, (text: string) => void>()
 
 const AGENT_TEMPLATE = (name: string) => `---
 description: what ${name} is for — the model reads this when it picks a subagent
-# model: provider/model        # run it on another of the server's models (default: the session's)
 # mode: plan                   # plan: read-only; edit; manual (asks you) — never looser than yours
 # tools: [read, glob, grep, list, bash, web_search, web_fetch]   # default: every tool a subagent may have
 # isolation: worktree          # work in a git checkout and branch of its own
@@ -96,23 +97,44 @@ report is all the session's model sees of your work.
 
 const TEAM_ROW = "\u0000team"
 
-/** The team's size for this session: a few usual pairs, or any typed as `<min> <max>`. */
+/** The Subagents setting as one range from 1 to the cap with two handles, Min and Max (harness
+ *  team.md §2): the numbers count this agent, so 1 is this agent alone. */
+function teamRange(c: CommandCtx): Option {
+  const t = c.app.team.team
+  const S = SUBAGENTS.setting
+  return {
+    label: S.label,
+    value: TEAM_ROW,
+    // The session's model cannot serve subagents itself: said where the size is set (subagents.md §4).
+    description: t.couldServe().own ? S.short : S.own,
+    range: { from: 1, to: t.cap, lo: t.min, hi: t.max, names: [S.min, S.max], onSubmit: (lo, hi) => applyTeamSize(c, `${lo} ${hi}`) },
+  }
+}
+
+const applyTeamSize = (c: CommandCtx, v: string) => (c.view.info(teamText(c.app, `size ${v}`)), c.refresh())
+
+/** The team's size for this session: the range, a few usual pairs, or any typed as `<min> <max>`. */
 function openTeamSize(c: CommandCtx) {
   const t = c.app.team.team
   const pairs = [
-    [1, 1, "one agent: no team, no helper"],
+    [1, 1, "this agent alone: no subagents"],
-    [1, 4, "you decide; up to three members"],
+    [1, 4, "this agent decides; up to three subagents"],
-    [2, 4, "a team from the start"],
+    [2, 4, "subagents from the start"],
-    [1, 8, "you decide; up to seven members"],
+    [1, 8, "this agent decides; up to seven subagents"],
-    [3, 8, "a team of at least three"],
+    [3, 8, "at least two subagents"],
-    [1, Math.min(16, t.cap), "a bigger team, with sub-managers"],
+    [1, Math.min(16, t.cap), "many subagents, with sub-managers"],
   ] as const
+  const S = SUBAGENTS.setting
   const seen = new Set<string>()
   const options = pairs
     .filter(([min, max]) => max <= t.cap && !seen.has(`${min} ${max}`) && seen.add(`${min} ${max}`))
-    .map(([min, max, what]) => ({ label: `${min} to ${max}`, value: `${min} ${max}`, description: what, current: t.min === min && t.max === max }))
+    .map(([min, max, what]) => ({ label: `${S.min} ${min}, ${S.max} ${max}`, value: `${min} ${max}`, description: what, current: t.min === min && t.max === max }))
-  const apply = (v: string) => (c.view.info(teamText(c.app, `size ${v}`)), c.refresh())
+  c.openDialog({
-  c.openDialog({ title: `Agents — now ${t.min} to ${t.max} (at most ${t.cap})`, options, onSelect: apply, custom: { label: "set to", onSubmit: apply } })
+    title: `${S.label} — ${S.min} ${t.min}, ${S.max} ${t.max} (1 to ${t.cap}, this agent included)`,
+    options: [teamRange(c), ...options],
+    onSelect: (v) => applyTeamSize(c, v),
+    custom: { label: "set to <min> <max>", onSubmit: (v) => applyTeamSize(c, v) },
+  })
 }
 
 export const COMMANDS: Command[] = [
@@ -146,7 +168,7 @@ export const COMMANDS: Command[] = [
       const describe = (a: (typeof list)[number]) =>
         [
           `${a.name} (${a.source}) — ${a.description}`,
-          `  model ${a.model ?? "the session's"} · mode ${a.mode ?? "the session's"} · tools ${a.tools?.join(", ") ?? "all a subagent may have"}${a.isolation ? " · own worktree" : ""}`,
+          `  model the session's, or one you name${a.model ? ` (its own model: ${a.model} is not used)` : ""} · mode ${a.mode ?? "the session's"} · tools ${a.tools?.join(", ") ?? "all a subagent may have"}${a.isolation ? " · own worktree" : ""}`,
         ].join("\n")
       if (named) return c.view.info(describe(named))
       c.view.info(
@@ -215,10 +237,10 @@ export const COMMANDS: Command[] = [
         c.setEffort(v as Level)
         c.view.info(control === "switch" ? `thinking ${v === "off" ? "off" : `on (${v})`}` : `effort: ${v}`)
       }
-      // The team's size sits beside the effort (team.md §2: shown only where a team is offered).
+      // The Subagents setting sits beside the effort (team.md §2, subagents.md §4: shown only where
-      const team = c.app.team.team
+      // subagents are offered).
-      const agentsRow = team.offered ? [{ label: `agents: ${team.min} to ${team.max}`, value: TEAM_ROW, description: "how many agents the team may have, you included" }] : []
+      const agentsRow = c.app.team.team.offered ? [teamRange(c)] : []
-      const choose = (v: string) => (v === TEAM_ROW ? openTeamSize(c) : set(v))
+      const choose = (v: string) => set(v)
       if (control === "none" && (arg || !agentsRow.length)) return c.view.info(`${m.ref} has one level (${levels[0]}): there is nothing to choose`)
       if (arg) return set(arg)
       if (control === "none") return c.openDialog({ title: `Chat settings — ${m.ref}`, options: agentsRow, onSelect: choose })
@@ -241,12 +263,29 @@ export const COMMANDS: Command[] = [
   },
   {
     name: "team",
-    args: "[size [<min> <max>] | stop [<name>]]",
+    args: "[size [<min> <max>] | stop [<name>] | open [<name>] | findings <name>]",
-    description: "the team: its agents and the room; its size; stop a member",
+    description: "subagents: the tree and the room; how many; stop one; their findings",
     run(c, arg) {
       // The tree and the room as the lead sees them — the person sees everything (team.md §5).
       const words = arg.trim().split(/\s+/).filter(Boolean)
       if (words[0] === "size" && words.length === 1) return openTeamSize(c)
+      // Spec 3.4.0 §11: a subagent's session, read-only, live or after it ended.
+      if (words[0] === "open") {
+        const t = c.app.team.team
+        const all = [...t.tree().filter((a) => a.manager !== null), ...t.ended]
+        const open = (key: string) => {
+          const a = t.find(key)
+          if (!a?.jobId) return c.view.info(`no subagent ${key} with a session here — /team lists them`, "warn")
+          c.enterJob(a.jobId)
+        }
+        if (words[1]) return open(words[1])
+        if (!all.length) return c.view.info("no subagents yet")
+        return c.openDialog({
+          title: "A subagent's session — read-only",
+          options: all.map((a) => ({ label: `${a.name} (${a.id})`, value: a.id, description: `${a.role} · ${TEAM.read.states[a.state]}` })),
+          onSelect: open,
+        })
+      }
       if (words[0] === "stop" && words.length === 1) {
         const members = c.app.team.team.tree().filter((a) => a.manager !== null)
         if (!members.length) return c.view.info("the team has no members")
@@ -687,8 +726,24 @@ export const COMMANDS: Command[] = [
   },
   {
     name: "context",
-    description: "where the context window goes",
+    args: "[<subagent>]",
-    run(c) {
+    description: "where the context window goes — the session's, or a subagent's",
+    run(c, arg) {
+      // Spec 3.4.0 §12: a subagent's context, by its name or id.
+      const who = arg.trim()
+      if (who) {
+        const t = c.app.team.team
+        const a = t.find(who)
+        if (!a) return c.view.info(`no subagent ${who} — /team lists them`, "warn")
+        const h = c.app.team.handle(a.name)
+        if (!h || a.endState) {
+          const u = a.context
+          return c.view.info(u ? `${a.name} (${a.id}): ${u.used} of ${u.window} tokens (${Math.floor((u.used * 100) / u.window)}%) when it ${a.endState ? TEAM.read.states[a.state] : "last said"}` : `${a.name} (${a.id}): no context use known yet`)
+        }
+        const se = h.engine
+        const sys = se.systemPrompt()
+        return c.view.info(`${a.name} (${a.id}) — ${a.role}\n${renderBreakdown(contextBreakdown(se, sys, [], undefined, se.o.tools))}`)
+      }
       const e = c.app.engine
       // The prompt the session sends (made now if no request has gone yet).
       const system = e.systemPrompt()
@@ -1006,7 +1061,7 @@ export const COMMANDS: Command[] = [
                       ? [...(e.def.values ?? [])]
                       : []
         c.openDialog({
-          title: `${key} — now ${show(e.value) || e.def.fallback || "unset"}${e.source === "default" ? "" : ` (${e.source})`}`,
+          title: `${key} — now ${show(e.value) || e.def.fallback || "unset"}${e.source === "default" ? "" : ` (${e.source})`}${e.def.hint ? ` · ${e.def.hint}` : ""}`,
           options: values.map((v) => ({ label: v, value: v, current: v === show(e.value) })),
           onSelect: (v) => askScope(key, v),
           custom: e.def.kind === "enum" || e.def.kind === "boolean" || e.def.kind === "effort" ? undefined : { label: "set to", onSubmit: (v) => askScope(key, v) },
@@ -1022,7 +1077,7 @@ export const COMMANDS: Command[] = [
         options: S.list().map((e) => ({
           label: e.def.key,
           value: e.def.key,
-          description: `${show(e.value) || e.def.fallback || "unset"}${e.source === "default" ? "" : ` · ${e.source}`}${e.def.reload ? " · after restart" : ""}`,
+          description: `${show(e.value) || e.def.fallback || "unset"}${e.source === "default" ? "" : ` · ${e.source}`}${e.def.reload ? " · after restart" : ""}${e.def.hint ? ` · ${e.def.hint}` : ""}`,
         })),
         onSelect: askValue,
       })

src/tui/components/dialog.tsx

@@ -8,6 +8,31 @@ export interface Option {
   value: string
   description?: string
   current?: boolean
+  /** A two-handle range on this row (the Subagents setting, harness team.md §2: one range from 1
+   *  to the cap, Min and Max). While the row is highlighted ←→ move the handle in focus and tab
+   *  moves the focus; Min never passes Max. Enter calls `onSubmit` with both, not `onSelect`. */
+  range?: Range
+}
+
+export interface Range {
+  /** The bounds of the track. */
+  from: number
+  to: number
+  /** The handles' values. */
+  lo: number
+  hi: number
+  /** The handles' names (Min, Max). */
+  names: [string, string]
+  onSubmit: (lo: number, hi: number) => void
+}
+
+/** The track as text: `─` outside the range, `━` inside it, `●` for a handle. */
+export function rangeTrack(r: Pick<Range, "from" | "to">, lo: number, hi: number, width = 16): { before: string; lo: string; between: string; hi: string; after: string } {
+  const span = Math.max(1, r.to - r.from)
+  const at = (v: number) => Math.round(((v - r.from) / span) * (width - 1))
+  const a = at(lo)
+  const b = Math.max(at(hi), a + (hi > lo ? 1 : 0))
+  return { before: "─".repeat(a), lo: "●", between: "━".repeat(Math.max(0, b - a - 1)), hi: b > a ? "●" : "", after: "─".repeat(Math.max(0, width - 1 - b)) }
 }
 
 export interface DialogSpec {
@@ -45,6 +70,16 @@ export function Dialog(p: { spec: DialogSpec; close: () => void }) {
     const hits = q ? p.spec.options.filter((o) => (o.label + " " + (o.description ?? "")).toLowerCase().includes(q)) : p.spec.options
     return p.spec.custom && query().trim() ? [{ label: `${p.spec.custom.label} "${query().trim()}"`, value: CUSTOM }, ...hits] : hits
   })
+  // A range row's handles, moved while the dialog is open and handed over on enter; esc drops them.
+  const [ranges, setRanges] = createSignal<Record<string, [number, number]>>({})
+  const [focus, setFocus] = createSignal<0 | 1>(1)
+  const valuesOf = (o: Option): [number, number] => ranges()[o.value] ?? [o.range!.lo, o.range!.hi]
+  const move = (o: Option, by: number) => {
+    const r = o.range!
+    const [lo, hi] = valuesOf(o)
+    const next: [number, number] = focus() === 0 ? [Math.min(hi, Math.max(r.from, lo + by)), hi] : [lo, Math.min(r.to, Math.max(lo, hi + by))]
+    setRanges((m) => ({ ...m, [o.value]: next }))
+  }
   const start = Math.max(0, p.spec.options.findIndex((o) => o.current))
   const [sel, setSel] = createSignal(start)
   // The spec is live: a list that redraws itself with fewer rows keeps the cursor on a row.
@@ -70,6 +105,8 @@ export function Dialog(p: { spec: DialogSpec; close: () => void }) {
         run(o.value)
       }
     } else if (key.name === "escape") p.close()
+    else if (shown()[sel()]?.range && (key.name === "left" || key.name === "right")) move(shown()[sel()]!, key.name === "left" ? -1 : 1)
+    else if (shown()[sel()]?.range && key.name === "tab") setFocus((f) => (f === 0 ? 1 : 0))
     else if (key.name === "up" || (key.ctrl && key.name === "p")) setSel((i) => (i + n - 1) % Math.max(n, 1))
     else if (key.name === "down" || (key.ctrl && key.name === "n")) setSel((i) => (i + 1) % Math.max(n, 1))
     else if (key.name === "return") {
@@ -79,8 +116,10 @@ export function Dialog(p: { spec: DialogSpec; close: () => void }) {
         const onSelect = p.spec.onSelect
         const custom = p.spec.custom
         const typed = query().trim()
+        const range = o.range && { submit: o.range.onSubmit, values: valuesOf(o) }
         p.close()
         if (o.value === CUSTOM) custom?.onSubmit(typed)
+        else if (range) range.submit(...range.values)
         else onSelect(o.value)
       }
     } else if (filterable() && key.name === "backspace") {
@@ -104,6 +143,30 @@ export function Dialog(p: { spec: DialogSpec; close: () => void }) {
         <For each={shown().slice(offset(), offset() + visible())}>
           {(o, i) => {
             const active = () => i() + offset() === sel()
+            if (o.range) {
+              const values = () => valuesOf(o)
+              const track = () => rangeTrack(o.range!, values()[0], values()[1])
+              const handle = (n: 0 | 1) => (active() && focus() === n ? c().accent : active() ? c().title : c().text)
+              return (
+                <box flexDirection="row" backgroundColor={active() ? c().menuCurrent : undefined}>
+                  <text wrapMode="none" truncate fg={active() ? c().title : c().text} attributes={active() ? TextAttributes.BOLD : 0}>
+                    {"  "}
+                    {o.label}
+                    {"  "}
+                    <span style={{ fg: handle(0) }}>{`${o.range.names[0]} ${values()[0]}`}</span>
+                    {" "}
+                    <span style={{ fg: c().dim }}>{track().before}</span>
+                    <span style={{ fg: handle(0) }}>{track().lo}</span>
+                    <span style={{ fg: c().accent }}>{track().between}</span>
+                    <span style={{ fg: handle(1) }}>{track().hi}</span>
+                    <span style={{ fg: c().dim }}>{track().after}</span>
+                    {" "}
+                    <span style={{ fg: handle(1) }}>{`${o.range.names[1]} ${values()[1]}`}</span>
+                    <span style={{ fg: c().muted }}>{o.description ? `  ${o.description}` : ""}</span>
+                  </text>
+                </box>
+              )
+            }
             return (
               <box flexDirection="row" backgroundColor={active() ? c().menuCurrent : undefined}>
                 <text wrapMode="none" truncate fg={active() ? c().title : c().text} attributes={active() ? TextAttributes.BOLD : 0}>
@@ -118,7 +181,11 @@ export function Dialog(p: { spec: DialogSpec; close: () => void }) {
         <Show when={shown().length === 0}>
           <text fg={c().muted}>nothing matches</text>
         </Show>
-        <text fg={c().dim}>{p.spec.hint ?? "↑↓ move · enter choose · esc close"}</text>
+        <text fg={c().dim}>
+          {shown()[sel()]?.range
+            ? `←→ move ${shown()[sel()]!.range!.names[focus()]} · tab ${shown()[sel()]!.range!.names[focus() === 0 ? 1 : 0]} · enter set · esc close`
+            : (p.spec.hint ?? "↑↓ move · enter choose · esc close")}
+        </text>
       </box>
     </box>
   )

src/tui/components/spinner.tsx

@@ -1,11 +1,18 @@
 import { createSignal, For, onCleanup } from "solid-js"
 import { duration } from "../../duration.ts"
+import { progressLine } from "../../session/progress.ts"
 import { useTui } from "../context.ts"
 import { fmtTokens } from "./status.tsx"
 
 /** "reading the prompt 64% · 90k of 140k · ~2m left": the server's own count (llama.cpp's
  *  return_progress). The time left is guessed from the reading so far, once there is enough of it;
  *  a long prompt reads slower towards its end, so it is a floor more than a promise. */
+/** How long the rest of the prompt will take, from the server's pace so far; "" until it is known. */
+export function etaOf(p: { total: number; done: number; processed: number; ms: number }): string {
+  if (p.processed >= 2048 && p.ms >= 2000 && p.done < p.total) return `~${duration(((p.total - p.done) / p.processed) * p.ms, true)} left`
+  return ""
+}
+
 export function reading(p: { total: number; done: number; processed: number; ms: number }): string {
   const pct = Math.floor((p.done / p.total) * 100)
   const parts = [`reading the prompt ${pct}%`, `${fmtTokens(p.done)} of ${fmtTokens(p.total)}`]
@@ -34,6 +41,16 @@ export function Spinner() {
   }
   const what = () => {
     const s = view.state
+    // Spec 3.4.0: the phase the engine says, drawn as every client draws it (progress.md §3).
+    if (s.now) {
+      tick()
+      // The CLI's own look around the shared line: how long the rest of the prompt will take.
+      // The server's counts as they come (the phase event is paced for the link; the screen is not).
+      const reading = s.now.phase === "reading_prompt" && s.progress
+      const now = reading ? { ...s.now, done: s.progress!.done, total: s.progress!.total } : s.now
+      const eta = reading ? etaOf(s.progress!) : ""
+      return `${progressLine(now)}${eta ? ` · ${eta}` : ""}`
+    }
     if (s.preparing) return `preparing ${s.preparing}`
     if (s.phase === "waiting" && s.progress) return reading(s.progress)
     if (s.phase === "thinking") return pick(sp().verbs, 30)
@@ -45,9 +62,9 @@ export function Spinner() {
     <box paddingLeft={1} height={1}>
       <text wrapMode="none" truncate>
         <span style={{ fg: theme().colors.accent }}>{pick(sp().frames, 1)} </span>
-        <span style={{ fg: theme().colors.muted }}>{what()}… </span>
+        <span style={{ fg: theme().colors.muted }}>{what()}{view.state.now ? " · " : "… "}</span>
         <span style={{ fg: theme().colors.dim }}>
-          {elapsed()} · {tokens()} · esc to stop
+          {view.state.now ? `turn ${elapsed()}` : elapsed()} · {tokens()} · esc to stop
         </span>
       </text>
     </box>

src/tui/components/subagent.tsx

@@ -27,7 +27,7 @@ export function SubagentView(p: { id: string; view: View; scrollRef?: (r: Scroll
     const j = job()
     if (!j) return `subagent · ${p.id} · gone · esc back`
     const state = j.state === "running" ? `running ${duration(Date.now() - j.started, true)}` : jobWords(j)
-    const message = j.subagent ? "type to message" : "cannot be messaged (an earlier process ran it)"
+    const message = j.team ? `read-only: ${j.team} works for the main agent` : j.subagent ? "type to message" : "cannot be messaged (an earlier process ran it)"
     return [`subagent ${j.agent ?? ""}`.trimEnd(), j.id, state, "esc back", message].join(" · ")
   }
   const what = () => job()?.description ?? ""

src/tui/state.ts

@@ -82,7 +82,7 @@ export interface Pending {
   /** What the model says the call is for. */
   purpose?: string
   /** The subagent (running as a job) that asks, when one does. */
-  source?: { jobId: string; agent: string; description: string }
+  source?: { jobId: string; agent: string; description: string; /** Spec 3.4.0: a subagent of the tree. */ subagent?: { id: string; name: string; position: string } }
   resolve: (r: AskReply) => void
 }
 
@@ -120,6 +120,8 @@ export interface ViewState {
   progress?: { total: number; done: number; processed: number; ms: number }
   /** The session's background jobs (the registry's list, kept from `job` events), running first. */
   jobs: TrackedJob[]
+  /** Spec 3.4.0 (harness/progress.md): what the agent is doing now — the spinner line. */
+  now?: import("../session/progress.ts").Progress | null
 }
 
 // What the model was given beside the words typed — attachments, and a session note
@@ -285,6 +287,9 @@ export function createView(bus: Bus) {
           if (e.state === "start") startTask()
           else inTask = false
           break
+        case "phase":
+          set("now", e.progress ? { ...e.progress } : null)
+          break
         case "sub_tool":
           set(
             "items",

tests/acp-jobs.test.ts

@@ -151,7 +151,7 @@ test("a background subagent: its transcript as history turns; send to a finished
   ], { route: (model) => (model.startsWith("g/") ? child : undefined) })
   config(fake.url, true)
   const { c, events, sid } = await client()
-  await c.request("session/prompt", { sessionId: sid, prompt: [{ type: "text", text: "go" }] })
+  await c.request("session/prompt", { sessionId: sid, prompt: [{ type: "text", text: "go (a subagent on g/m)" }] })
   expect(await until(() => events.some((e) => e.type === "job" && e.state === "end"))).toBe(true)
   const listed: any = await c.request("_lembas/jobs", { sessionId: sid })
   expect(listed.jobs[0]).toMatchObject({ id: "job1", kind: "subagent", state: "done", agent: "explore", description: "find the bug", model: "g/m" })

tests/acp.test.ts

@@ -268,7 +268,7 @@ test("the terminal: a real shell through the link, only where remote.terminal sa
   config(fake.url)
   const root = project()
   const off = await client({ roots: [root], maxMode: "edit", approvalTimeoutMs: 50, requireTrust: true })
-  expect(off.init._meta.lembas.capabilities).toEqual(["directories", "configure", "steer", "sessions", "status", "delete", "compact", "title", "files", "commands", "attachments", "ask", "plan", "model", "turns", "context", "library", "prototypes", "plan_state", "rewind", "jobs", "show"])
+  expect(off.init._meta.lembas.capabilities).toEqual(["directories", "configure", "steer", "sessions", "status", "delete", "compact", "title", "files", "commands", "attachments", "ask", "plan", "model", "turns", "context", "library", "prototypes", "plan_state", "rewind", "jobs", "show", "progress", "tree", "agents"])
   await expect(off.c.request("_lembas/terminal/open", { cwd: root })).rejects.toThrow("remote.terminal is off")
 
   const { c, init } = await client({ roots: [root], maxMode: "edit", approvalTimeoutMs: 50, requireTrust: true, terminal: true })

tests/capacity.test.ts

@@ -51,10 +51,37 @@ test("task on a model the server cannot hold beside the session's is refused, an
   )
   writeFileSync(join(paths.config, "config.yaml"), "model: f/m\n")
   const a = createApp({ cwd: mkdtempSync(join(tmpdir(), "ph-cap-")), mode: "edit", store: false, asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) } })
-  await a.engine.prompt("get a second opinion")
+  await a.engine.prompt("get a second opinion from f/other")
   // Two requests: the session's, then its next step — never one for f/other.
   expect(fake.requests.map((r: any) => r.model)).toEqual(["f/m", "f/m"])
   const result = fake.requests[1].messages.at(-1).content as string
-  expect(result).toContain("holds one model at a time")
+  expect(result).toContain("That model is on the same one-model server as this chat's model; name a model on another server.")
-  expect(result).toContain("Do this part yourself")
+})
+
+test("spec 3.3.0: task on another model the person did not name is refused, and nothing runs", async () => {
+  fake = fakeProvider([
+    { chunks: [toolCall(0, "t1", "task", JSON.stringify({ description: "second opinion", prompt: "look at it", model: "f/other" }))] },
+    { chunks: [delta({ content: "I will do it myself." })] },
+  ])
+  mkdirSync(paths.config, { recursive: true })
+  writeFileSync(join(paths.config, "connections.yaml"), `connections:\n  f:\n    type: webui\n    url: ${fake.url}\n    models: { f/m: {}, f/other: {} }\n`, { mode: 0o600 })
+  writeFileSync(join(paths.config, "config.yaml"), "model: f/m\n")
+  const a = createApp({ cwd: mkdtempSync(join(tmpdir(), "ph-cap-")), mode: "edit", store: false, asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) } })
+  await a.engine.prompt("get a second opinion")
+  expect(fake.requests.map((r: any) => r.model)).toEqual(["f/m", "f/m"])
+  expect(fake.requests[1].messages.at(-1).content as string).toContain("Use another model for subagents only when the person names it.")
+})
+
+test("spec 3.3.0: a one-request session model is not its own subagent, and nothing else is picked", async () => {
+  fake = fakeProvider([
+    { chunks: [toolCall(0, "t1", "task", JSON.stringify({ description: "look", prompt: "look at it" }))] },
+    { chunks: [delta({ content: "I will do it myself." })] },
+  ])
+  mkdirSync(paths.config, { recursive: true })
+  writeFileSync(join(paths.config, "connections.yaml"), `connections:\n  f:\n    type: webui\n    url: ${fake.url}\n    models: { f/m: { parallel: 1 }, f/other: {} }\n`, { mode: 0o600 })
+  writeFileSync(join(paths.config, "config.yaml"), "model: f/m\n")
+  const a = createApp({ cwd: mkdtempSync(join(tmpdir(), "ph-cap-")), mode: "edit", store: false, asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) } })
+  await a.engine.prompt("look into it")
+  expect(fake.requests.map((r: any) => r.model)).toEqual(["f/m", "f/m"])
+  expect(fake.requests[1].messages.at(-1).content as string).toContain("This model serves one request at a time: name another model, on another server, for subagents.")
 })

tests/conformance.ts

@@ -27,9 +27,13 @@ import { jobNote, type NoteCase } from "../src/jobs/note.ts"
 import { compactAsk, noteFor, offeredFor, toldAt, unavailable, type Now, type Requirement, type Told } from "../src/session/prefix.ts"
 import { envLines, promptText } from "../src/prompt/assemble.ts"
 import { absPath, type ToolContext } from "../src/tool/tool.ts"
-import { TOOL_SPECS } from "../src/harness.ts"
+import { LOOP, TOOL_SPECS } from "../src/harness.ts"
-import { Team, TEAM_TOOLS, type Capacity } from "../src/team/team.ts"
+import { LEAD_TOOLS, memberTools, Team, TEAM_TOOLS, type Capacity } from "../src/team/team.ts"
 import { noteBlock } from "../src/session/prefix.ts"
+import { offer, subagentModel, type ModelInfo } from "../src/team/subagents.ts"
+import { progressLine, progressText, type Progress } from "../src/session/progress.ts"
+import { checkMode, requestRefusal, tighten, whoDecides, type Grant } from "../src/team/delegation.ts"
+import { fillPrompt as fillText } from "../src/prompt/fill.ts"
 
 type Case = Record<string, unknown>
 
@@ -233,12 +237,35 @@ export const AREAS: Record<string, (c: Case) => unknown> = {
     if (c.op === "resolve") return absPath(c.path as string, { root: (c.root as string).replace(/\/+$/, ""), cwd: c.started as string } as ToolContext)
     throw new Error(`paths: unknown op ${String(c.op)}`)
   },
+  // Subagents (subagents.md, spec 3.3.0): the model a subagent runs on, through the rules spawn and
+  // team_add ask; the offer through the Team the app makes (its tools, setting and note).
+  subagents: (c) => {
+    const models = Object.entries(c.models as Record<string, Omit<ModelInfo, "ref">>).map(([ref, m]) => ({ ref, ...m }))
+    const byRef = (ref: string) => models.find((m) => m.ref === ref)!
+    const session = byRef(c.session as string)
+    if (c.op === "model") {
+      const r = subagentModel(session, c.model === undefined ? undefined : byRef(c.model as string), (c.said as string[] | undefined) ?? [])
+      return "refused" in r ? { refused: r.refused } : { model: r.model }
+    }
+    if (c.op === "offer") {
+      const subagents = c.subagents !== false
+      const pure = offer(session, models, subagents)
+      const team = new Team({ min: 1, max: 4, subagents, model: session.ref, capacity: (ref) => byRef(ref).capacity, models: () => models.map((m) => m.ref) })
+      if (team.offered !== pure.offered) throw new Error("subagents: the Team and the rule disagree on the offer")
+      const noted = team.offered && team.noteLines()[0]!.split("\n").includes(LOOP.context.note.team_own)
+      return { offered: team.offered, own: pure.own, note: noted ? LOOP.context.note.team_own : null }
+    }
+    throw new Error(`subagents: unknown op ${String(c.op)}`)
+  },
   // Teams (team.md): one lead session's team, step by step, through the CLI's own rules
   // (team/team.ts); the app drives the same class from real turns.
   team: (c) => {
-    const start = c.start as { min: number; max: number; subagents?: boolean; capacity: Capacity }
+    const start = c.start as { min: number; max: number; subagents?: boolean; capacity: Capacity; mode?: Mode }
     let capacity = start.capacity
-    const team = new Team({ min: start.min, max: start.max, subagents: start.subagents, model: "m", capacity: () => capacity })
+    const team = new Team({ min: start.min, max: start.max, subagents: start.subagents, model: "m", capacity: () => capacity, mode: start.mode ?? "edit", steps: { main: 200, sub: 50 } })
+    // What each agent's turn was started with (a brief, or the messages that woke it).
+    const started: Record<string, string> = {}
+    team.effects = { start: (n, t) => void (started[n] = t), deliver: () => {}, stop: () => {} }
     const delivered = () => {
       const out: Record<string, string[]> = {}
       for (const d of team.delivered.splice(0)) (out[d.name] ??= []).push(d.text)
@@ -249,8 +276,8 @@ export const AREAS: Record<string, (c: Case) => unknown> = {
       switch (st.op) {
         case "offered": {
           const a = team.get(by)!
-          const tools = !team.offered ? [] : a.leads ? [...TEAM_TOOLS] : ["team_read", "team_send"]
+          const tools = !team.offered ? [] : by === "lead" ? [...LEAD_TOOLS] : memberTools(a)
-          return { tools: tools.sort(), setting: team.offered }
+          return { tools: [...tools].sort(), setting: team.offered }
         }
         case "call": {
           const args = st.args as Record<string, any>
@@ -260,20 +287,29 @@ export const AREAS: Record<string, (c: Case) => unknown> = {
             const room = no ? [] : team.readRoom(by, args.since ?? 0).messages.map((m) => ({ n: m.n, from: m.from, to: m.to.join(", ") }))
             return { room, delivered: delivered() }
           }
-          if (st.tool === "team_add") result = team.add(by, args as any).text
+          const first = (TEAM_TOOLS as readonly string[]).includes(st.tool) ? team.refusal(by, st.tool, args) : undefined
+          if (first) result = first
+          else if (st.tool === "team_add") result = team.add(by, args as any).text
           else if (st.tool === "team_send") result = team.send(by, args as any)
           else if (st.tool === "team_stop") result = team.stopCmd(by, args.name)
+          else if (st.tool === "team_assign") result = team.assign(by, args as any)
+          else if (st.tool === "team_replace") result = team.replace(by, args as any).text
+          else if (st.tool === "team_finish") result = team.finishTask(by, args.report)
+          else if (st.tool === "team_whoami") result = team.whoami(by)
+          else if (st.tool === "team_findings") result = args.action === "write" ? team.noteFinding(by, args.text ?? "") : team.readFindings(by, args.of)
+          else if (st.tool === "team_request") result = team.request(by, args as any, TOOL_SPECS[args.tool]?.risk as any)
+          else if (st.tool === "team_decide") result = team.decide(by, args as any)
           else {
             result = team.refusal(by, st.tool, args) ?? null
-            if (st.tool === "task" && result === null && team.offered) team.helperStart(by)
+            if (st.tool === "task" && result === null && team.offered) team.taskStart(by)
           }
           return { result, delivered: delivered() }
         }
         case "finish":
           team.finish(by, st.reply ?? "", st.reason ?? "stop")
           return { delivered: delivered() }
-        case "helper_done":
+        case "task_done":
-          team.helperDoneBy(by)
+          team.taskDoneBy(by)
           return null
         case "set":
           if (st.capacity) capacity = st.capacity
@@ -289,10 +325,65 @@ export const AREAS: Record<string, (c: Case) => unknown> = {
         }
         case "state":
           return team.state()
+        // Spec 3.4.0.
+        case "context":
+          team.setContext(by, st.used, st.window)
+          return null
+        case "context_full":
+          team.pause(by)
+          return { delivered: delivered() }
+        case "started":
+          return { text: started[by] ?? null }
+        case "tree": {
+          const t = team.info()
+          const row = (a: (typeof t.agents)[number]) => ({ id: a.id, name: a.name, parent: a.parent, state: a.state, mode: a.mode, request: a.request })
+          return { size: t.size, open_slots: t.open_slots, agents: t.agents.map(row), ended: t.ended.map((a) => ({ id: a.id, name: a.name, state: a.state })) }
+        }
+        case "main_mode":
+          team.setMainMode(st.mode)
+          return null
       }
       throw new Error(`team: unknown op ${String(st.op)}`)
     })
   },
+  // Progress (progress.md, spec 3.4.0): the text and the line a client draws.
+  progress: (c) => {
+    const p = c.progress as Progress
+    return { text: progressText(p, c.now as number), line: progressLine(p, c.now as number) }
+  },
+  // Delegation (subagents.md §7, spec 3.4.0): the mode check, who decides a call, team_request's
+  // check, and tightening down the tree.
+  delegation: (c) => {
+    if (c.op === "add") {
+      const r = checkMode(c.parent_mode as Mode, c.mode as Mode | undefined, c.request as string[] | undefined)
+      return "error" in r ? { result: r.error, mode: null } : { result: null, mode: r.mode }
+    }
+    if (c.op === "call") {
+      const call = c.call as { tool: string; line?: string; path?: string }
+      const req: PermissionRequest =
+        call.tool === "bash"
+          ? { permission: "bash", class: "execute", patterns: [call.line!], command: call.line!, paths: ["/project"] }
+          : { permission: call.tool === "read" ? "read" : "edit", class: call.tool === "read" ? "read" : "write", patterns: [call.path!], paths: [`/project/${call.path}`] }
+      const v = whoDecides(req, {
+        ctx: { rules: toRules(DEFAULT_RULES, "default"), hardline: BUILTIN_HARDLINE, root: "/project", planDir: "/project/.agent/plans", projectDir: "/project/.agent" },
+        mode: c.mode as Mode,
+        main: c.main_mode as Mode,
+        request: (c.request as string[] | undefined) ?? [],
+        grants: (c.grants as Grant[] | undefined) ?? [],
+        unattended: c.unattended === true,
+      })
+      // The text is compared unless the case says the reason is each project's own.
+      return v.answer === "refused" && c.own_reason !== true ? { answer: v.answer, text: v.text } : { answer: v.answer }
+    }
+    if (c.op === "request") {
+      const cls = TOOL_SPECS[c.tool as string]?.risk as "read" | "write" | "execute" | undefined
+      return { result: requestRefusal(c.tool as string, cls, c.request as string[]) ?? fillText(LOOP.team.texts.requested, { id: `${LOOP.team.request.id_prefix}1` }) }
+    }
+    if (c.op === "switch") {
+      return tighten(c.main_to as Mode, c.agents as { name: string; parent: string; mode: Mode }[])
+    }
+    throw new Error(`delegation: unknown op ${String(c.op)}`)
+  },
   library_scope: (c) => resolveBase(c.bases as { id: string; name: string }[], c.base as string | null),
   toolcall_index: (c) => {
     // The deltas as the OpenAI stream carries them: `function.name` and `function.arguments`.

tests/context.test.ts

@@ -87,7 +87,8 @@ describe("context", () => {
     await a.engine.prompt("read it")
     const system = a.engine.compose(a.engine.mode)
     const text = renderBreakdown(contextBreakdown(a.engine, system, [], undefined, a.engine.o.tools))
-    expect(text).toContain("window 32.8k")
+    expect(text).toContain("/ 32.8k tokens")
+    expect(text).toMatch(/until auto-compact \(at 85%\) — \/compact to compact now/)
     expect(text).toContain("system prompt")
     expect(text).toMatch(/tool definitions \(\d+\)/)
     expect(text).toMatch(/tool results\s+514/) // one 6000-character line, cut at 2000 by read, under its header

tests/e2e.test.ts

@@ -54,7 +54,7 @@ describe("end to end", () => {
     expect(fake!.requests[0].messages[0].content).toContain("Permission mode: edit")
     // knowledge_search/get are offered once any knowledge base has documents, and the suite shares
     // one LEMBAS_HOME: another file's base must not decide this list (found in the full run).
-    expect(fake!.requests[0].tools.map((t: any) => t.function.name).filter((n: string) => !n.startsWith("knowledge_"))).toEqual(["read", "write", "edit", "multiedit", "apply_patch", "glob", "grep", "list", "bash", "job_output", "job_list", "job_stop", "job_send", "web_search", "web_fetch", "task", "team_add", "team_send", "team_read", "team_stop", "plan_write", "plan_update", "ask_user", "plan_submit", "memory", "session_search", "notes_search", "note_view", "note_manage", "skills_list", "skill_view", "skill_manage", "settings", "recall"])
+    expect(fake!.requests[0].tools.map((t: any) => t.function.name).filter((n: string) => !n.startsWith("knowledge_"))).toEqual(["read", "write", "edit", "multiedit", "apply_patch", "glob", "grep", "list", "bash", "job_output", "job_list", "job_stop", "job_send", "web_search", "web_fetch", "task", "team_add", "team_send", "team_read", "team_stop", "team_whoami", "team_findings", "team_decide", "team_assign", "team_replace", "plan_write", "plan_update", "ask_user", "plan_submit", "memory", "session_search", "notes_search", "note_view", "note_manage", "skills_list", "skill_view", "skill_manage", "settings", "recall"])
     const ends = events.filter((e) => e.type === "tool_end")
     expect(ends.map((e) => e.type === "tool_end" && e.result.title)).toEqual(["hello.txt · 1–1 of 1", "hello.txt +1 −1"])
     const sid = store.sessions(1)[0]!.id

tests/harness.test.ts

@@ -153,7 +153,7 @@ test("spec 3.0.0: every tool says where it is offered; the CLI's are the agent's
   // plan_write's former name rather than a tool.
   expect(TOOL_SPECS.todo).toBeUndefined()
   expect(TOOL_SPECS.plan_write!.aliases.cli).toEqual(["todo"])
-  const shared = ["ask_user", "edit", "glob", "grep", "job_list", "job_output", "job_send", "job_stop", "list", "memory", "plan_submit", "plan_update", "plan_write", "read", "recall", "report", "session_search", "show", "skill_manage", "skill_view", "skills_list", "task", "team_add", "team_read", "team_send", "team_stop", "web_fetch", "web_search", "write"]
+  const shared = ["ask_user", "edit", "glob", "grep", "job_list", "job_output", "job_send", "job_stop", "list", "memory", "plan_submit", "plan_update", "plan_write", "read", "recall", "report", "session_search", "show", "skill_manage", "skill_view", "skills_list", "task", "team_add", "team_assign", "team_decide", "team_findings", "team_finish", "team_read", "team_replace", "team_request", "team_send", "team_stop", "team_whoami", "web_fetch", "web_search", "write"]
   expect(Object.values(TOOL_SPECS).filter((t) => t.scope === "shared").map((t) => t.name).sort()).toEqual(shared.sort())
 })
 

tests/lembas-login.test.ts

@@ -555,3 +555,17 @@ test("lembas login --token-stdin --server --json: the token from stdin, the devi
   expect((await cli(["login", "--token-stdin", fake.url], "lmb_secret")).code).toBe(2)
   expect((await cli(["login", "--server", fake.url], "lmb_secret")).code).toBe(2)
 })
+
+test("lembas login over a sign-in the instance no longer accepts signs in afresh, not in a loop", async () => {
+  fake = lembas()
+  await login(fake.url, { ...quick, onCode: () => {} })
+  // The instance forgot the token kept here (revoked, or another instance's database).
+  fake.token = "lmb_fresh"
+  fake.polls = 0
+  const r = await cli(["login", fake.url, "--json"], "")
+  expect(r.code).toBe(0)
+  const lines = r.out.trim().split("\n").map((l) => JSON.parse(l))
+  expect(lines[0]).toMatchObject({ event: "code" })
+  expect(lines.at(-1)).toMatchObject({ event: "done", fresh: true, connection: "example" })
+  expect(readFileSync(keyFile("example"), "utf8").trim()).toBe("lmb_fresh")
+})

tests/subagent.test.ts

@@ -197,7 +197,7 @@ describe("background subagents (spec 2.0.0)", () => {
       ],
       [{ chunks: [toolCall(0, "r1", "read", '{"path":"calc.py"}')] }, { chunks: [delta({ content: "calc.py:2 divides by len(v) - 1." })] }],
     )
-    await a.turns.prompt("look into it in the background")
+    await a.turns.prompt("look into it in the background (subagents on g/m or g/n)")
     const started = fake!.requests[1].messages.at(-1).content as string
     expect(started).toStartWith("[#0001]\nStarted subagent explore as job1 (find the bug).")
     await settled(a, 3)
@@ -233,7 +233,7 @@ describe("background subagents (spec 2.0.0)", () => {
         { chunks: [delta({ content: "Second report." })] },
       ],
     )
-    await a.turns.prompt("go")
+    await a.turns.prompt("go (subagents on g/m or g/n)")
     expect(fake!.requests[2].messages.at(-1).content).toContain("Sent to job1")
     await settled(a, 6)
     const child = other!.requests
@@ -252,7 +252,7 @@ describe("background subagents (spec 2.0.0)", () => {
       [{ chunks: [toolCall(0, "t1", "task", JSON.stringify({ description: "find loaders", prompt: "Find the loaders", model: "g/m" }))] }, { chunks: [delta({ content: "Thanks." })] }],
       [{ chunks: [toolCall(0, "p1", "report", JSON.stringify({ text: "There are two loaders; which one?" }))] }, { chunks: [delta({ content: "Final: the YAML one." })] }],
     )
-    await a.turns.prompt("go")
+    await a.turns.prompt("go (subagents on g/m or g/n)")
     const msgs = fake!.requests[1].messages
     const note = msgs.find((m: any) => m.role === "user" && String(m.content).startsWith(MARKS.message))
     expect(note?.content).toContain("job1 (explore: find loaders):\nThere are two loaders; which one?")
@@ -278,7 +278,7 @@ describe("background subagents (spec 2.0.0)", () => {
       [{ chunks: [toolCall(0, "b1", "bash", JSON.stringify({ command: "true", description: "nothing" }))] }, { chunks: [delta({ content: "ran" })] }],
       { asker },
     )
-    await a.turns.prompt("go")
+    await a.turns.prompt("go (subagents on g/m or g/n)")
     expect(seen).toEqual([{ jobId: "job1", agent: "general", description: "run it" }])
   })
 
@@ -297,7 +297,7 @@ describe("background subagents (spec 2.0.0)", () => {
       [{ chunks: [delta({ content: "slow" })], gapMs: 400 }],
       { oneAtATime: true },
     )
-    await a.turns.prompt("go")
+    await a.turns.prompt("go (subagents on g/m or g/n)")
     const results = fake!.requests[1].messages.filter((m: any) => m.role === "tool").map((m: any) => m.content as string)
     expect(results[0]).toStartWith("[#0001]\nStarted subagent explore as job1")
     expect(results[1]).toContain("holds one model at a time, and a subagent is already running on g/m there")
@@ -330,7 +330,7 @@ test("a cancelled worktree subagent keeps its branch, and the result names it",
   const root = a.project.root
   for (const args of [["init", "-q", "-b", "main"], ["config", "user.email", "t@e"], ["config", "user.name", "T"], ["config", "commit.gpgsign", "false"], ["add", "calc.py"], ["commit", "-q", "-m", "init"]]) git(root, args)
   ;(a.project as any).gitRoot = root
-  const turn = a.turns.prompt("fix it in a worktree")
+  const turn = a.turns.prompt("fix it in a worktree (subagents on g/m or g/n)")
   const until = Date.now() + 8000
   while (Date.now() < until && !events.some((e) => e.type === "sub_tool" && e.name === "write")) await Bun.sleep(20)
   const write = events.find((e) => e.type === "sub_tool" && e.name === "write")

tests/subagents-v2.test.ts

@@ -0,0 +1,207 @@
+// Subagents v2 (harness spec 3.4.0, subagents.md §5–§14; progress.md) through the whole app: a
+// subagent's team_finish ends its turn, a narrowed one is refused what only the main agent may
+// allow and asks for it, a call nobody in the tree can allow reaches the person naming the
+// subagent, a subagent whose context fills gets one final turn and ends, every engine says its
+// progress as phase events, and an agent chat's tree goes over the link. The rules themselves are
+// conformance/team.json's and delegation.json's.
+import { afterEach, describe, expect, test } from "bun:test"
+import { mkdirSync, mkdtempSync, realpathSync, writeFileSync } from "node:fs"
+import { tmpdir } from "node:os"
+import { join } from "node:path"
+import { createApp, type App } from "../src/app.ts"
+import type { AskReply, Event } from "../src/bus/index.ts"
+import { paths } from "../src/config/paths.ts"
+import { LOOP } from "../src/harness.ts"
+import { AcpAgent } from "../src/acp/agent.ts"
+import { Peer, type Transport } from "../src/acp/rpc.ts"
+import { setTrust } from "../src/project/root.ts"
+import { delta, fakeProvider, toolCall, type Fake, type Scripted } from "./fake-provider.ts"
+
+let fake: Fake | undefined
+let app: App | undefined
+afterEach(async () => {
+  await app?.close()
+  app = undefined
+  fake?.stop()
+})
+
+const T = LOOP.team
+const say = (text: string): Scripted => ({ chunks: [delta({ content: text }, "stop")] })
+const call = (id: string, name: string, args: unknown, usage?: { prompt_tokens: number; completion_tokens: number }): Scripted => ({
+  chunks: [toolCall(0, id, name, JSON.stringify(args)), delta({}, "tool_calls"), ...(usage ? [{ choices: [], usage }] : [])],
+})
+const system = (body: any): string => body.messages[0].content
+const memberName = (body: any): string | undefined => /You are "([a-z0-9-]+)" \(s\d+\), a subagent/.exec(system(body))?.[1]
+const lastUser = (body: any): string => {
+  const u = [...body.messages].reverse().find((m: any) => m.role === "user")
+  return typeof u.content === "string" ? u.content : u.content.map((p: any) => p.text ?? "").join("")
+}
+const toolResults = (body: any): string[] => body.messages.filter((m: any) => m.role === "tool").map((m: any) => (typeof m.content === "string" ? m.content : JSON.stringify(m.content)).replace(/^\[#\d+\]\n/, ""))
+
+function setup(answer: (body: any) => Scripted | undefined, spec = "{ context: 32768, parallel: 4 }") {
+  fake = fakeProvider(answer)
+  mkdirSync(paths.config, { recursive: true })
+  writeFileSync(join(paths.config, "connections.yaml"), `connections:\n  fake:\n    type: webui\n    url: ${fake.url}\n    models:\n      fake/m: ${spec}\n`, { mode: 0o600 })
+  writeFileSync(join(paths.config, "config.yaml"), "model: fake/m\n")
+  const dir = realpathSync(mkdtempSync(join(tmpdir(), "lembas-sub2-")))
+  Bun.spawnSync(["git", "init", "-q", dir])
+  setTrust(dir, "trusted")
+  return dir
+}
+
+async function until(what: () => boolean, ms = 10000) {
+  const end = Date.now() + ms
+  while (!what()) {
+    if (Date.now() > end) throw new Error("timed out")
+    await Bun.sleep(20)
+  }
+}
+
+/** The lead's script: the calls in order, then a closing line; each member's by its name. */
+function scripted(lead: Scripted[], members: Record<string, Scripted[]>) {
+  const at: Record<string, number> = {}
+  return (body: any): Scripted => {
+    const who = memberName(body) ?? "lead"
+    const list = who === "lead" ? lead : (members[who] ?? [])
+    const n = (at[who] = (at[who] ?? -1) + 1)
+    return list[n] ?? say(`${who} is done.`)
+  }
+}
+
+describe("subagents v2 (spec 3.4.0)", () => {
+  test("team_finish reports to the parent and ends the subagent: nothing runs after that call", async () => {
+    const dir = setup(
+      scripted([call("a1", "team_add", { name: "api", role: "the API", task: "Write the API.", goal: "a service" }), say("api is on it.")], {
+        api: [call("f1", "team_finish", { report: "The API is done: src/api.ts." }), call("x1", "write", { path: "late.txt", content: "x" })],
+      }),
+    )
+    app = createApp({ cwd: dir, mode: "edit", asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) }, store: false })
+    await app.engine.prompt("build it")
+    await until(() => app!.team.team.ended.length === 1)
+    await until(() => !app!.engine.busy)
+    const t = app.team.team
+    expect(t.ended[0]).toMatchObject({ id: "s1", name: "api", state: "done" })
+    expect(t.room.find((m) => m.report)?.text).toBe("The API is done: src/api.ts.")
+    // Its one request: nothing after the team_finish call.
+    expect(fake!.requests.filter((r) => memberName(r) === "api")).toHaveLength(1)
+    expect(await Bun.file(join(dir, "late.txt")).exists()).toBe(false)
+  }, 20000)
+
+  test("permissions: a narrowed subagent is told to ask the main agent; a call nobody in the tree can allow reaches the person, naming the subagent", async () => {
+    const asked: any[] = []
+    const dir = setup(
+      scripted(
+        [
+          call("a1", "team_add", { name: "docs", role: "the docs", task: "Write the docs.", goal: "docs", mode: "manual", request: ["write"] }),
+          call("a2", "team_add", { name: "ops", role: "the release", task: "Push it." }),
+          say("Both are on it."),
+        ],
+        {
+          docs: [call("w1", "write", { path: "README.md", content: "# Docs\n" }), say("asked")],
+          ops: [call("b1", "bash", { command: "git push origin main" }), say("pushed or not")],
+        },
+      ),
+    )
+    app = createApp({ cwd: dir, mode: "edit", asker: { ask: async (req): Promise<AskReply> => (asked.push(req), { kind: "deny" }) }, store: false })
+    await app.engine.prompt("go")
+    await until(() => fake!.requests.filter((r) => memberName(r) === "docs").length >= 2 && fake!.requests.filter((r) => memberName(r) === "ops").length >= 2)
+    const docs = fake!.requests.filter((r) => memberName(r) === "docs")
+    expect(toolResults(docs[1])[0]).toBe(T.texts.requestable)
+    // The docs subagent has team_request (it has request options); the ops subagent does not.
+    expect(docs[0].tools.map((x: any) => x.function.name)).toContain("team_request")
+    const ops = fake!.requests.filter((r) => memberName(r) === "ops")
+    expect(ops[0].tools.map((x: any) => x.function.name)).not.toContain("team_request")
+    expect(ops[0].tools.map((x: any) => x.function.name)).not.toContain("team_add")
+    // An all-edit tree: the subagent's push reaches the person, not the main agent.
+    expect(asked.map((r) => [r.tool, r.source?.subagent?.name])).toEqual([["bash", "ops"]])
+    expect(asked[0].source.subagent).toEqual({ id: "s2", name: "ops", position: "the release" })
+    expect(await Bun.file(join(dir, "README.md")).exists()).toBe(false)
+  }, 20000)
+
+  test("a subagent whose context fills is paused, gets one final turn, and ends; its parent gets the notice and its last reply", async () => {
+    const dir = setup(
+      scripted([call("a1", "team_add", { name: "scan", role: "scans the code", task: "Scan everything.", goal: "a map" }), say("scan is on it.")], {
+        scan: [
+          call("n1", "team_findings", { action: "write", text: "src/ has 12 modules." }, { prompt_tokens: 3700, completion_tokens: 20 }),
+          say("Saved my findings; the tests are left."),
+        ],
+      }),
+      "{ context: 4096, parallel: 4 }",
+    )
+    app = createApp({ cwd: dir, mode: "edit", asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) }, store: false })
+    await app.engine.prompt("map it")
+    await until(() => app!.team.team.ended.length === 1)
+    const t = app.team.team
+    expect(t.ended[0]).toMatchObject({ name: "scan", state: "out_of_context" })
+    expect(t.ended[0]!.findings.map((f) => f.text)).toEqual(["src/ has 12 modules."])
+    const scans = fake!.requests.filter((r) => memberName(r) === "scan")
+    // The final turn: its user message is the notice, under its mark.
+    expect(lastUser(scans[1])).toContain(T.texts.context_full.replace("{{steps}}", String(T.context.final_steps)))
+    const report = t.room.find((m) => m.report)!
+    expect(report.text).toBe(`${T.texts.gone.replace("{{name}}", "scan").replace("{{id}}", "s1").replace("{{name}}", "scan")}\n\nSaved my findings; the tests are left.`)
+  }, 20000)
+
+  test("every engine says its progress: reading the prompt, a tool, writing, and nothing once the turn is done", async () => {
+    const dir = setup(() => undefined)
+    let n = 0
+    fake!.stop()
+    fake = fakeProvider(() => (n++ === 0 ? call("l1", "list", {}) : say("Listed.")))
+    writeFileSync(join(paths.config, "connections.yaml"), `connections:\n  fake:\n    type: webui\n    url: ${fake.url}\n    models:\n      fake/m: { context: 32768 }\n`, { mode: 0o600 })
+    app = createApp({ cwd: dir, mode: "edit", asker: { ask: async (): Promise<AskReply> => ({ kind: "once" }) }, store: false })
+    const phases: (string | null)[] = []
+    app.bus.on((e: Event) => e.type === "phase" && phases.push(e.progress?.phase ?? null))
+    await app.engine.prompt("list it")
+    expect(phases).toContain("reading_prompt")
+    expect(phases).toContain("running_tool")
+    expect(phases).toContain("writing")
+    expect(phases.at(-1)).toBeNull()
+    expect(app.engine.progress.current).toBeNull()
+  }, 20000)
+})
+
+function pair(): [Transport, Transport] {
+  const make = () => ({ msg: (_: string) => {}, end: () => {} })
+  const a = make()
+  const b = make()
+  const side = (me: typeof a, o: typeof a): Transport => ({
+    send: (t) => queueMicrotask(() => o.msg(t)),
+    onMessage: (fn) => (me.msg = fn),
+    onClose: (fn) => (me.end = fn),
+    close: () => (me.end(), o.end()),
+  })
+  return [side(a, b), side(b, a)]
+}
+
+test("an agent chat's tree over the link: the setting comes down with the turn; the tree, progress, findings and transcript come up", async () => {
+  const dir = setup(
+    scripted([call("a1", "team_add", { name: "api", role: "the API", task: "Write the API.", goal: "a service" }), say("api is on it.")], {
+      api: [call("n1", "team_findings", { action: "write", text: "Login is POST /session." }), say("API done.")],
+    }),
+  )
+  const [ta, tb] = pair()
+  new AcpAgent(new Peer(ta))
+  const c = new Peer(tb)
+  const events: any[] = []
+  c.on("_lembas/event", (p) => void events.push(p.event))
+  c.handle("session/request_permission", () => ({ outcome: { outcome: "selected", optionId: "once" } }))
+  const init: any = await c.request("initialize", { protocolVersion: 1, clientCapabilities: { _meta: { lembas: { protocol: 3 } } } })
+  expect(init._meta.lembas.capabilities).toEqual(expect.arrayContaining(["progress", "tree", "agents"]))
+  const s: any = await c.request("session/new", { cwd: dir, mcpServers: [], _meta: { lembas: { agents: { min: 1, max: 3 }, said: ["Use the main model only."] } } })
+  await c.request("session/prompt", { sessionId: s.sessionId, prompt: [{ type: "text", text: "build it" }], _meta: { lembas: { agents: { min: 1, max: 3 } } } })
+  await until(() => events.some((e) => e.type === "team" && e.tree?.agents.some((a: any) => a.name === "api" && a.state === "finished")))
+  const view: any = await c.request("_lembas/team", { sessionId: s.sessionId })
+  expect(view.tree).toMatchObject({ min: 1, max: 3, size: 2 })
+  expect(view.tree.agents.map((a: any) => [a.id, a.name, a.parent])).toEqual([
+    ["lead", "lead", null],
+    ["s1", "api", "lead"],
+  ])
+  expect(view.room.map((m: any) => [m.from, m.kind])).toEqual([["api", "report"]])
+  expect(await c.request("_lembas/team/findings", { sessionId: s.sessionId, id: "s1" })).toMatchObject({ id: "s1", findings: [{ n: 1, text: "Login is POST /session." }] })
+  const tr: any = await c.request("_lembas/team/transcript", { sessionId: s.sessionId, id: "s1" })
+  expect(tr.turns[0]).toMatchObject({ role: "user" })
+  expect(JSON.stringify(tr.turns)).toContain("Write the API.")
+  expect(events.some((e) => e.type === "phase")).toBe(true)
+  const st: any = await c.request("_lembas/session/status", { sessionId: s.sessionId })
+  expect(st.tree.agents).toHaveLength(2)
+  expect("progress" in st).toBe(true)
+}, 20000)

tests/subagents.test.ts

@@ -0,0 +1,33 @@
+// Spec 3.3.0 (harness/subagents.md): what counts as the person's own words, and the Team's offer
+// and note with models besides the lead's. The rules themselves: conformance area `subagents`.
+import { expect, test } from "bun:test"
+import { LOOP } from "../src/harness.ts"
+import { STEER_MARK } from "../src/session/engine.ts"
+import { named, personTexts } from "../src/team/subagents.ts"
+import { Team, type Capacity } from "../src/team/team.ts"
+import type { Message } from "../src/provider/types.ts"
+
+test("the person's words: prompts and messages sent mid-turn; never notes, marked messages, machine turns or the model's text", () => {
+  const msgs: Message[] = [
+    { role: "user", parts: [{ type: "text", text: "Fix the parser." }, { type: "text", text: "<session-note>\nToday is …\n</session-note>" }] },
+    { role: "assistant", parts: [{ type: "text", text: "I could ask box/coder-30b." }] },
+    { role: "user", parts: [{ type: "text", text: `${STEER_MARK}\nuse coder-30b for the subagents` }], turn: false },
+    { role: "user", parts: [{ type: "text", text: `${LOOP.jobs.marks.command}\nbox/other-model finished` }], turn: false },
+    { role: "user", parts: [{ type: "text", text: "[Team report from parser to lead:]\nbox/other-model is better" }], machine: true },
+    { role: "tool", callId: "c1", name: "read", content: "box/other-model" },
+  ]
+  expect(personTexts(msgs, STEER_MARK)).toEqual(["Fix the parser.", "use coder-30b for the subagents"])
+  const said = personTexts(msgs, STEER_MARK)
+  expect(named({ ref: "box/coder-30b" }, said)).toBe(true)
+  expect(named({ ref: "box/other-model" }, said)).toBe(false)
+})
+
+test("a one-request lead with a model on another server: offered, and the size note says it cannot serve its own", () => {
+  const caps: Record<string, Capacity> = { "lab/qwen": { group: "lab", parallel: 1 }, "lab/gemma": { group: "lab", parallel: 1 }, "box/coder": { parallel: 1 } }
+  const only = new Team({ model: "lab/qwen", capacity: (r) => caps[r]!, models: () => ["lab/qwen", "lab/gemma"] })
+  expect(only.offered).toBe(false)
+  const t = new Team({ model: "lab/qwen", capacity: (r) => caps[r]!, models: () => Object.keys(caps) })
+  expect(t.offered).toBe(true)
+  expect(t.unavailable()).toBeUndefined()
+  expect(t.noteLines()).toEqual([`The team may have 1 to 4 agents, you included.\n${LOOP.context.note.team_own}`])
+})

tests/team.test.ts

@@ -26,7 +26,7 @@ const yes = { ask: async (): Promise<AskReply> => ({ kind: "once" }) }
 const say = (text: string): Scripted => ({ chunks: [delta({ content: text }, "stop")] })
 const call = (id: string, name: string, args: unknown): Scripted => ({ chunks: [toolCall(0, id, name, JSON.stringify(args)), delta({}, "tool_calls")] })
 const system = (body: any): string => body.messages[0].content
-const isMember = (body: any, name: string) => system(body).includes(`You are "${name}", a member of a team`)
+const isMember = (body: any, name: string) => new RegExp(`You are "${name}" \\(s\\d+\\), a subagent`).test(system(body))
 const lastUser = (body: any): string => {
   const u = [...body.messages].reverse().find((m: any) => m.role === "user")
   return typeof u.content === "string" ? u.content : u.content.map((p: any) => p.text ?? "").join("")
@@ -94,8 +94,8 @@ describe("a team at work (team.md)", () => {
     expect(t.room.map((m) => ({ from: m.from, to: m.to, report: m.report }))).toEqual([{ from: "parser", to: ["lead"], report: true }])
     expect(t.get("parser")!.state).toBe("finished")
     expect(t.isManager("lead")).toBe(true)
-    expect(teamLabel(t)).toBe("team 2/4")
+    expect(teamLabel(t)).toBe("agents 2/4")
-    expect(teamText(app, "")).toContain("- parser — the parser · finished")
+    expect(teamText(app, "")).toContain("- parser (s1) — the parser · finished")
     expect(await Bun.file(join(dir, "x.txt")).exists()).toBe(false)
     // A member is not a job to the model.
     expect(app.jobs().every((j) => j.team === "parser")).toBe(true)
@@ -135,10 +135,10 @@ describe("a team at work (team.md)", () => {
     const dir = setup(() => say(`ok ${n++}`))
     app = createApp({ cwd: dir, mode: "edit", asker: yes, store: false })
     await app.engine.prompt("one")
-    expect(teamText(app, "size 3 2")).toContain("1 ≤ min ≤ max ≤ 32")
+    expect(teamText(app, "size 3 2")).toContain("1 ≤ Min ≤ Max ≤ 32")
-    expect(teamText(app, "size 1 40")).toContain("1 ≤ min ≤ max ≤ 32")
+    expect(teamText(app, "size 1 40")).toContain("1 ≤ Min ≤ Max ≤ 32")
     expect([app.team.team.min, app.team.team.max]).toEqual([1, 4])
-    expect(teamText(app, "size 2 6")).toContain("2 to 6")
+    expect(teamText(app, "size 2 6")).toContain("Min 2, Max 6")
     expect(app.settings.get("agents.max").value).toBe(6)
     await app.engine.prompt("two")
     expect(lastUser(fake!.requests[1])).toContain("The team may have 2 to 6 agents, you included.")

tests/tui/app.test.tsx

@@ -372,14 +372,14 @@ test("thinking is one line — its text is not shown; the provider's count of it
   expect(await t.until("SECRET-INNER-STEP")).toBe(true)
 })
 
-test("while it thinks: the thought says only Thinking…; the line at the bottom has the verb, the task's time and tokens", async () => {
+test("while it thinks: the thought says only Thinking…; the line at the bottom has the phase (progress.md), the task's time and tokens", async () => {
   t = await tui([{ gapMs: 1500, chunks: [delta({ reasoning_content: "Weighing it up: HIDDEN-WHILE-THINKING and more words here." }), delta({ content: "Done thinking." })] }])
   await t.input.typeText("go")
   t.input.pressEnter()
   expect(await t.until("~15 tokens")).toBe(true)
   const frame = t.frame()
   expect(frame).toMatch(/^\s*Thinking\.{1,3}\s*$/m)
-  expect(frame).toMatch(/(pondering|contemplating|musing|cogitating|ruminating|deliberating|mulling|reflecting|reasoning|analyzing|synthesizing|formulating)… \d+s · ~15 tokens · esc to stop/)
+  expect(frame).toMatch(/Thinking · \d+s · turn \d+s · ~15 tokens · esc to stop/)
   expect(frame).not.toContain("HIDDEN-WHILE-THINKING")
   expect(await t.until("Done thinking.")).toBe(true)
   expect(t.frame()).toMatch(/Thought for \d\.\ds · ~15 tokens/)
@@ -397,7 +397,7 @@ test("the line at the bottom counts the whole task: the clock and the tokens go
   // The spinner line redraws on its own tick, slower than the reply's text: on a slow machine the
   // frame that first shows "Half" can still carry the line from before. The 1.2 s gap holds this
   // state, so wait for the line within it rather than reading the first frame.
-  const counted = /writing… \d+s · ~4\d tokens · esc to stop/
+  const counted = /Writing(: [\d.]+ tok\/s)? · \d+s · turn \d+s · ~4\d tokens · esc to stop/
   for (let i = 0; i < 20 && !counted.test(t.frame()); i++) await Bun.sleep(50), await t.settle()
   expect(t.frame()).toMatch(counted)
   expect(await t.until("way.")).toBe(true)
@@ -922,10 +922,12 @@ test("while the server reads a long prompt: the line at the bottom says how far
   t = await tui([{ gapMs: 500, chunks: [pp(0, 0), pp(80000, 160000), delta({ content: "Read it all." }, "stop")] }])
   await t.input.typeText("go")
   t.input.pressEnter()
-  expect(await t.until("reading the prompt 64% · 90k of 140k · ~1m 40s left")).toBe(true)
+  // At 100 columns the line is cut in its middle; its start and its end are what is checked.
+  expect(await t.until("~1m 40s left")).toBe(true)
+  expect(t.frame()).toContain("Reading the prompt: 90000 / 140000 tokens (64%")
   expect(await t.until("Read it all.")).toBe(true)
   await t.settle()
-  expect(t.frame()).not.toContain("reading the prompt")
+  expect(t.frame()).not.toContain("Reading the prompt")
 })
 
 test("a mouse wheel that arrives as a burst of ↑ scrolls the transcript; one ↑ still walks history", async () => {

tests/tui/dialog.test.tsx

@@ -36,3 +36,50 @@ test("dialog: arrow down + enter picks the second option — with a spec that cl
   expect(picked).toEqual(["b"])
   expect(closed).toBe(1)
 })
+
+test("dialog: a range row is one track with two handles — ←→ move the one in focus, tab the other, enter hands both over", async () => {
+  const set: [number, number][] = []
+  const picked: string[] = []
+  const ctx = { theme: () => DEFAULT_THEME, dims: () => ({ width: 72, height: 16 }) } as never
+  const [spec, setSpec] = createSignal<DialogSpec | undefined>({
+    title: "Effort",
+    options: [
+      { label: "high", value: "high", current: true },
+      { label: "Subagents", value: "team", description: "1 is this agent alone", range: { from: 1, to: 32, lo: 1, hi: 4, names: ["Min", "Max"], onSubmit: (lo, hi) => set.push([lo, hi]) } },
+    ],
+    onSelect: (v) => picked.push(v),
+  })
+  setup = await testRender(
+    () => (
+      <TuiContext.Provider value={ctx}>
+        <Show when={spec()}>
+          <Dialog spec={spec()!} close={() => setSpec(undefined)} />
+        </Show>
+      </TuiContext.Provider>
+    ),
+    { width: 72, height: 16 },
+  )
+  await setup.renderOnce()
+  expect(setup.captureCharFrame()).toContain("Subagents  Min 1 ●")
+  setup.mockInput.pressArrow("down")
+  await setup.renderOnce()
+  expect(setup.captureCharFrame()).toContain("←→ move Max · tab Min")
+  setup.mockInput.pressArrow("right")
+  setup.mockInput.pressArrow("right")
+  await setup.renderOnce()
+  expect(setup.captureCharFrame()).toContain("Max 6")
+  setup.mockInput.pressTab()
+  setup.mockInput.pressArrow("right")
+  setup.mockInput.pressArrow("right")
+  setup.mockInput.pressArrow("left")
+  await setup.renderOnce()
+  expect(setup.captureCharFrame()).toContain("Min 2")
+  // Min never passes Max, and Min never goes under 1.
+  for (let i = 0; i < 10; i++) setup.mockInput.pressArrow("right")
+  await setup.renderOnce()
+  expect(setup.captureCharFrame()).toContain("Min 6")
+  setup.mockInput.pressEnter()
+  await setup.renderOnce()
+  expect(set).toEqual([[6, 6]])
+  expect(picked).toEqual([])
+})