Resident subagents
V0.x breaking change: zut no longer supports subprocess subagent state.
Previous state is ignored rather than read or migrated; remove it when you
no longer need it. Existing child jobs are never resumed.
Subagents are independent core.Agent conversations resident in the same zut
process as their parent. A child has a stable, private session identity, its own
provider client and tool registry, a cancellation boundary, and an authoritative
journal. Starting a child never launches another zut executable.
Profiles
Profiles are Markdown files discovered from ~/.agents/agents/*.md and paths
in ZUT_AGENT_PROFILES. They are inputs only; zut does not execute them.
---
name: reviewer
description: Read-only code reviewer
tools: [read, grep]
model: openai-codex/gpt-5.6-luna
thinking: high
systemPromptMode: replace
inheritProjectContext: false
inheritSkills: false
fastMode: false
---
Inspect the requested scope and report evidence-backed findings.
tools is a capability boundary. An omitted value inherits the child-safe
catalogue, tools: [] grants no tools, and an explicit list replaces the
default. Names unavailable to the child (including host-only extensions) and
tools denied by subagents.allowed_tools are omitted from an explicit list;
they never grant the child additional access. model, provider, thinking,
the prompt mode, project/skill inheritance, and fastMode are resolved before
the child is accepted. Spawn-level model, provider, reasoning, and fast-mode
values take precedence where supplied.
Children are always fresh conversations: they receive the delegated task and
resolved host/profile context, never the parent transcript or hidden reasoning.
They cannot recursively create more children.
Lifecycle and persistence
The manager records each accepted child under its managed state root:
subagents/<child-id>/
transcript.jsonl # authoritative accepted turns and finalized messages
owner.lock # host-ownership coordination; do not modify or remove
.transcript-backup-* # retained private pre-repair transcript, when repaired
metadata.json # rebuildable bounded state
result.json # latest state plus a bounded final assistant summary
patch.diff # optional worktree capture
Acceptance is durable before subagent_spawn reports success. Finalized user,
assistant, tool-call, and tool-result messages are journaled; streamed deltas
and hidden reasoning are not exposed as ordinary history. A tool call is stored
before execution and reconciliation repairs incomplete call/result pairs.
All children stop when the host exits. On the next start, queued or running
turns are marked interrupted; zut never replays their task. Resume only with
an explicit new prompt through subagent_resume, the child-session composer,
or /subagents resume <id> <prompt>.
Multiple zut processes may share a state root. Each resident journal has one
host owner at a time. A process that finds a child owned elsewhere reports it
as such and does not inspect, recover, or modify its transcript. Resume that
child from its owning host or after the owner exits. On a subsequent safe
reconciliation, zut repairs only the known legacy false-recovery sequence; it
leaves any ambiguous journal corruption untouched. A repair retains the
pre-repair transcript as .transcript-backup-*; it can contain private session
data and is intentionally not removed automatically.
subagent_status reports owned_elsewhere: true when another process owns a
child. The field is omitted when false. A foreign-owned child cannot be resumed,
or have its history or result read, from this process.
The global scheduler admits the oldest eligible accepted prompt, runs at most
one turn for each child, and defaults to six concurrent turns. Set
subagents.max_concurrent to a positive limit to override it. Queueing is
unlimited by default; a positive subagents.queue_timeout cancels an accepted
prompt that has not received a slot and records its terminal failure durably.
subagents.allowed_tools is an allowlist for child-visible tools and
subagents.allowed_roots limits eligible child workspaces. Each child receives
a cumulative weighted-token budget equal to its resolved model's context window.
Removed legacy settings, including tui_subagent_position and
subagents.budget_ratio, are ignored. Resident journals created before this
budget policy are incompatible: zut leaves them untouched, reports a recovery
error, and requires a new child rather than resuming them.
Budget accounting charges uncached input, cache writes, and output fully; cache
reads count at 25%. It is separate from active context-window usage and survives
follow-up turns and compaction. The TUI shows the child's current budget usage,
but does not alter its instructions, tools, or output cap before exhaustion.
Providers report exact input usage only after a response, so the terminal
response can make displayed usage exceed 100%; no later request is sent. An
exhausted child cannot accept follow-up turns; spawn a replacement when more
work is required.
required: true makes a delegated result an obligation of the parent turn.
Failed, cancelled, and interrupted required work remains unresolved until an
explicit successful follow-up. A host restart never retries it automatically.
Delegation ownership
Interactive proactive delegation keeps the primary agent on the critical path.
Before spawning, it selects useful local work and delegates only a concrete,
bounded sidecar with a non-overlapping question, responsibility, package, or
file set. Immediate blockers and tightly coupled work stay local. While a child
is active, it owns that scope; the primary continues only its preselected
independent work instead of repeating the investigation, review, testing, or
implementation. If an explicit user request or active workflow hands blocking
work to a child, the primary yields for the completion update. Once a worktree
child succeeds, the primary may apply and verify its captured patch without
redoing the delegated task.
Explicit headless --orchestrate remains manager-only. Its primary divides the
request into disjoint worker scopes, coordinates active children, and yields
rather than implementing or reproducing delegated work. Children do not inherit
the parent's strict orchestrator role. Headless orchestration waits for every
accepted child before continuing its completion wave; required: true
additionally creates a durable obligation. When proactive
interactive delegation cannot spawn because of launch-time policy, the primary
continues locally; a strict headless orchestrator instead reports the missing
required capability.
The model-facing tools retain their logical names:
subagent_spawn accepts task, optional agent, model and provider,
reasoning, fast_mode, required, wait, and isolation (shared or
worktree).
wait is an explicit whole-second value from 1 through 300; when omitted,
spawning returns immediately. A timed-out wait leaves the
accepted child active, whether it is queued or running. Do not retry that
task until it reaches a terminal failure, cancellation, or interruption. It
returns a logical subagent://<id> reference.subagent_status returns bounded current state for one child or the current
set immediately.subagent_stop stops one live child.subagent_resume accepts an explicit follow-up prompt for an existing child.
Budget-exhausted children are terminal and reject follow-ups.
Child execution started by subagent_spawn is asynchronous unless it receives
an explicit wait value. For an unwaited spawn, completion arrives through the
host’s typed completion update; subagent_status only returns the current
state immediately. Do not use sleep loops, repeated status calls, journal
files, or terminal UI inspection as a completion signal.
Successful completions include the final visible assistant summary, capped at
256 KiB; open the child session for the complete durable transcript.
/subagents opens the resident-child list, ordered by the latest state change.
Each row leads with the profile (or model), shows a human terminal state such
as completed, a relative update time, and a short ID. Press Down from the
main composer to open this list without clearing a draft when a child exists.
Use arrows and Enter to open a child session; /subagents new <task> creates a child; /subagents logs <id>
opens its history; /subagents result <id> shows the bounded final summary;
kill <id> stops it; and resume <id> <prompt> continues it.
The child session uses the same transcript renderer as the main session.
Recent history is bounded and loaded asynchronously; PgUp asks for an older
page. Up/Down preserves the current reading position, and PgDn returns to the
live tail after unread updates accumulate. The screen combines durable finalized
history with an immutable in-memory projection of the active turn. Enter submits
the local composer only after the manager durably accepts it; Esc closes the session. History may contain task
text, tool arguments, tool results, and source-derived output, so treat it as
private local session data.
Worktrees and references
Shared children use the host working directory. With isolation: "worktree",
zut prepares a detached Git worktree, captures a patch and changed-file list on
successful completion, and never applies that patch automatically. Failed or
interrupted worktrees are retained for inspection; an idle successful worktree
is cleaned when its child is explicitly stopped.
Logical references avoid exposing runtime paths:
subagent://<id>
subagent://<id>/history
subagent://<id>/result
subagent://<id>/patch
Provider caching
Each main and child conversation sends a stable logical session ID and a new
turn ID per accepted prompt to the provider layer. Provider adapters decide
whether and how to use them: OpenAI/Codex translates a supported stable session
into its cache key and routing header, while other adapters preserve deterministic
request input and use only cache controls their provider/model/endpoint supports.
An OpenAI-compatible wire shape is not itself evidence that a cache extension
or transport feature is available. Models continue through their declared
provider transport; a transport change must not change the logical session or
turn identity used by the provider request.
source
1# Resident subagents
2
3> **V0.x breaking change:** zut no longer supports subprocess subagent state.
4> Previous state is ignored rather than read or migrated; remove it when you
5> no longer need it. Existing child jobs are never resumed.
6
7Subagents are independent `core.Agent` conversations resident in the same zut
8process as their parent. A child has a stable, private session identity, its own
9provider client and tool registry, a cancellation boundary, and an authoritative
10journal. Starting a child never launches another zut executable.
11
12## Profiles
13
14Profiles are Markdown files discovered from `~/.agents/agents/*.md` and paths
15in `ZUT_AGENT_PROFILES`. They are inputs only; zut does not execute them.
16
17```markdown
18---
19name: reviewer
20description: Read-only code reviewer
21tools: [read, grep]
22model: openai-codex/gpt-5.6-luna
23thinking: high
24systemPromptMode: replace
25inheritProjectContext: false
26inheritSkills: false
27fastMode: false
28---
29
30Inspect the requested scope and report evidence-backed findings.
31```
32
33`tools` is a capability boundary. An omitted value inherits the child-safe
34catalogue, `tools: []` grants no tools, and an explicit list replaces the
35default. Names unavailable to the child (including host-only extensions) and
36tools denied by `subagents.allowed_tools` are omitted from an explicit list;
37they never grant the child additional access. `model`, `provider`, `thinking`,
38the prompt mode, project/skill inheritance, and `fastMode` are resolved before
39the child is accepted. Spawn-level model, provider, reasoning, and fast-mode
40values take precedence where supplied.
41
42Children are always fresh conversations: they receive the delegated task and
43resolved host/profile context, never the parent transcript or hidden reasoning.
44They cannot recursively create more children.
45
46## Lifecycle and persistence
47
48The manager records each accepted child under its managed state root:
49
50```text
51subagents/<child-id>/
52 transcript.jsonl # authoritative accepted turns and finalized messages
53 owner.lock # host-ownership coordination; do not modify or remove
54 .transcript-backup-* # retained private pre-repair transcript, when repaired
55 metadata.json # rebuildable bounded state
56 result.json # latest state plus a bounded final assistant summary
57 patch.diff # optional worktree capture
58```
59
60Acceptance is durable before `subagent_spawn` reports success. Finalized user,
61assistant, tool-call, and tool-result messages are journaled; streamed deltas
62and hidden reasoning are not exposed as ordinary history. A tool call is stored
63before execution and reconciliation repairs incomplete call/result pairs.
64
65All children stop when the host exits. On the next start, queued or running
66turns are marked `interrupted`; zut never replays their task. Resume only with
67an explicit new prompt through `subagent_resume`, the child-session composer,
68or `/subagents resume <id> <prompt>`.
69
70Multiple zut processes may share a state root. Each resident journal has one
71host owner at a time. A process that finds a child owned elsewhere reports it
72as such and does not inspect, recover, or modify its transcript. Resume that
73child from its owning host or after the owner exits. On a subsequent safe
74reconciliation, zut repairs only the known legacy false-recovery sequence; it
75leaves any ambiguous journal corruption untouched. A repair retains the
76pre-repair transcript as `.transcript-backup-*`; it can contain private session
77data and is intentionally not removed automatically.
78
79`subagent_status` reports `owned_elsewhere: true` when another process owns a
80child. The field is omitted when false. A foreign-owned child cannot be resumed,
81or have its history or result read, from this process.
82
83The global scheduler admits the oldest eligible accepted prompt, runs at most
84one turn for each child, and defaults to six concurrent turns. Set
85`subagents.max_concurrent` to a positive limit to override it. Queueing is
86unlimited by default; a positive `subagents.queue_timeout` cancels an accepted
87prompt that has not received a slot and records its terminal failure durably.
88`subagents.allowed_tools` is an allowlist for child-visible tools and
89`subagents.allowed_roots` limits eligible child workspaces. Each child receives
90a cumulative weighted-token budget equal to its resolved model's context window.
91Removed legacy settings, including `tui_subagent_position` and
92`subagents.budget_ratio`, are ignored. Resident journals created before this
93budget policy are incompatible: zut leaves them untouched, reports a recovery
94error, and requires a new child rather than resuming them.
95
96Budget accounting charges uncached input, cache writes, and output fully; cache
97reads count at 25%. It is separate from active context-window usage and survives
98follow-up turns and compaction. The TUI shows the child's current budget usage,
99but does not alter its instructions, tools, or output cap before exhaustion.
100Providers report exact input usage only after a response, so the terminal
101response can make displayed usage exceed 100%; no later request is sent. An
102exhausted child cannot accept follow-up turns; spawn a replacement when more
103work is required.
104
105`required: true` makes a delegated result an obligation of the parent turn.
106Failed, cancelled, and interrupted required work remains unresolved until an
107explicit successful follow-up. A host restart never retries it automatically.
108
109## Delegation ownership
110
111Interactive proactive delegation keeps the primary agent on the critical path.
112Before spawning, it selects useful local work and delegates only a concrete,
113bounded sidecar with a non-overlapping question, responsibility, package, or
114file set. Immediate blockers and tightly coupled work stay local. While a child
115is active, it owns that scope; the primary continues only its preselected
116independent work instead of repeating the investigation, review, testing, or
117implementation. If an explicit user request or active workflow hands blocking
118work to a child, the primary yields for the completion update. Once a worktree
119child succeeds, the primary may apply and verify its captured patch without
120redoing the delegated task.
121
122Explicit headless `--orchestrate` remains manager-only. Its primary divides the
123request into disjoint worker scopes, coordinates active children, and yields
124rather than implementing or reproducing delegated work. Children do not inherit
125the parent's strict orchestrator role. Headless orchestration waits for every
126accepted child before continuing its completion wave; `required: true`
127additionally creates a durable obligation. When proactive
128interactive delegation cannot spawn because of launch-time policy, the primary
129continues locally; a strict headless orchestrator instead reports the missing
130required capability.
131
132## Tools and slash commands
133
134The model-facing tools retain their logical names:
135
136- `subagent_spawn` accepts `task`, optional `agent`, `model` and `provider`,
137 `reasoning`, `fast_mode`, `required`, `wait`, and `isolation` (`shared` or
138 `worktree`).
139 `wait` is an explicit whole-second value from 1 through 300; when omitted,
140 spawning returns immediately. A timed-out wait leaves the
141 accepted child active, whether it is queued or running. Do not retry that
142 task until it reaches a terminal failure, cancellation, or interruption. It
143 returns a logical `subagent://<id>` reference.
144- `subagent_status` returns bounded current state for one child or the current
145 set immediately.
146- `subagent_stop` stops one live child.
147- `subagent_resume` accepts an explicit follow-up prompt for an existing child.
148 Budget-exhausted children are terminal and reject follow-ups.
149
150Child execution started by `subagent_spawn` is asynchronous unless it receives
151an explicit `wait` value. For an unwaited spawn, completion arrives through the
152host’s typed completion update; `subagent_status` only returns the current
153state immediately. Do not use sleep loops, repeated status calls, journal
154files, or terminal UI inspection as a completion signal.
155Successful completions include the final visible assistant summary, capped at
156256 KiB; open the child session for the complete durable transcript.
157
158`/subagents` opens the resident-child list, ordered by the latest state change.
159Each row leads with the profile (or model), shows a human terminal state such
160as `completed`, a relative update time, and a short ID. Press Down from the
161main composer to open this list without clearing a draft when a child exists.
162Use arrows and Enter to open a child session; `/subagents new <task>` creates a child; `/subagents logs <id>`
163opens its history; `/subagents result <id>` shows the bounded final summary;
164`kill <id>` stops it; and `resume <id> <prompt>` continues it.
165
166The child session uses the same transcript renderer as the main session.
167Recent history is bounded and loaded asynchronously; PgUp asks for an older
168page. Up/Down preserves the current reading position, and PgDn returns to the
169live tail after unread updates accumulate. The screen combines durable finalized
170history with an immutable in-memory projection of the active turn. Enter submits
171the local composer only after the manager durably accepts it; Esc closes the session. History may contain task
172text, tool arguments, tool results, and source-derived output, so treat it as
173private local session data.
174
175## Worktrees and references
176
177Shared children use the host working directory. With `isolation: "worktree"`,
178zut prepares a detached Git worktree, captures a patch and changed-file list on
179successful completion, and never applies that patch automatically. Failed or
180interrupted worktrees are retained for inspection; an idle successful worktree
181is cleaned when its child is explicitly stopped.
182
183Logical references avoid exposing runtime paths:
184
185```text
186subagent://<id>
187subagent://<id>/history
188subagent://<id>/result
189subagent://<id>/patch
190```
191
192## Provider caching
193
194Each main and child conversation sends a stable logical session ID and a new
195turn ID per accepted prompt to the provider layer. Provider adapters decide
196whether and how to use them: OpenAI/Codex translates a supported stable session
197into its cache key and routing header, while other adapters preserve deterministic
198request input and use only cache controls their provider/model/endpoint supports.
199An OpenAI-compatible wire shape is not itself evidence that a cache extension
200or transport feature is available. Models continue through their declared
201provider transport; a transport change must not change the logical session or
202turn identity used by the provider request.