Commit Graph
39 Commits
Author SHA1 Message Date
Dudu-0223 6cc45f22bd feat: 在内置配置中关闭 subagent 和 Team 的 interrupt 工具 (#7986)
* feat(subagent): disable interrupt tool by default

* test(subagent): update remaining default tool inventories

* fix: disable agent interruption in shipped configuration

* test: configure interruption explicitly in custom preset fixture

* test: update tool count in collapse display snapshots

* docs: clarify agent interruption opt-in and verify preset controls

* test: cover macOS CUA channel policy forwarding on every platform

* test: move CUA coverage repair to a separate master change
2026-11-07 06:55:25 +00:00
_Kerman beb5186e36 Merge origin/master into xtr/runtime-context-reminders 2026-11-05 15:23:36 +08:00
Dudu-0223 ba4b6d69ea refactor(subagent): remove shared directory and scope guidance 2026-11-04 16:59:07 +08:00
Dudu-0223 bb3ac8e304 Merge master into feat/subagent-catalog-names 2026-11-04 13:51:31 +08:00
_Kerman 9776b567bc Merge remote-tracking branch 'origin/master' into xtr/runtime-context-reminders 2026-11-04 13:26:15 +08:00
_Kerman efa57ca366 Merge origin/master into xtr/runtime-context-reminders 2026-11-04 13:26:14 +08:00
Dudu-0223 a6f6b98053 Merge master into feat/subagent-catalog-names 2026-11-04 12:04:15 +08:00
sjw1231 1d5f9f7607 Merge remote-tracking branch 'origin/master' into feat/apply-patch-row-summary 2026-11-04 12:03:48 +08:00
Dudu-0223 fd8dfa270d test(subagent): refresh naming snapshots against current base 2026-11-03 16:02:27 +08:00
Dudu-0223 0d18d256c3 Merge master and simplify wait_agent input guidance 2026-11-03 14:51:55 +08:00
Dudu-0223 8223beef55 docs(subagent): explain wait input delivery after code execution 2026-11-03 14:14:56 +08:00
Dudu-0223 9f57925000 Merge remote-tracking branch 'origin/master' into feat/subagent-catalog-names
# Conflicts:
#	docs/persistence-schema.json
#	docs/tool-catalog.i18n.yaml
2026-11-03 13:49:18 +08:00
Dudu-0223 0cb47b9d03 fix(subagent): integrate root-call scoped wait protection
Integrate PR #7774 with one-shot code guidance and matching snapshots. Cover distinct nested call IDs sharing a root and concurrent model calls with pending input.
2026-11-03 13:29:33 +08:00
Dudu-0223 c3fda8ed51 docs(subagent): retain wait_agent input examples 2026-11-03 12:16:55 +08:00
imccyu a3f6f94c1e test(snapshot): refresh fixtures for the 10-query web_search default 2026-11-03 12:13:05 +08:00
Dudu-0223 1f522605ed fix(subagent): clarify one-shot waits from code 2026-11-03 12:06:10 +08:00
Dudu-0223 a4c933ad1f fix(subagent): fail unmatched name paths and tighten name admission
Review round on the naming change.

`resolveName` now distinguishes an address from a Session id: a
slash-separated address never is one, so an unmatched path fails with
AGENT_NAME_NOT_FOUND instead of reaching member lookup as an unknown id,
while a single token keeps the Session-id fallback. Path resolution also
treats external and unknown children as leaves, so a name below an external
run reports the same clear failure instead of surfacing a session-query error
for an identity that owns no local Session.

Native V4 admission rejects a repeated sibling name in own catalog facts,
symmetrically with its duplicate-child check, so a log that would make name
lookup ambiguous fails before any member is resolved.

`SiblingNameReservations.release()` checks that it still owns the map entry,
so a stale or repeated release cannot drop a newer reservation for the same
creator.

`list_agents` explains corrupt, unsupported, and unavailable: they report that
the row's own descendants could not be read, and the row remains a member the
caller can message. The earlier wording restricted messaging to running or
inactive rows, which excluded addressable members.

`isAgentName` stays module-local; the package entry exports only
ROOT_AGENT_NAME, matching its current callers.

Package READMEs, the subsystem page, the Agent Note, and the generated API
catalogue follow; recorded-session, CLI, and Web expectations are refreshed.
2026-11-02 22:43:01 +08:00
Dudu-0223 6644a234a5 Merge master into feat/subagent-catalog-names 2026-11-02 17:46:04 +08:00
Chinesezjc ec587a015d test(snapshot): carry the read guidance into the apply-patch scenarios
The apply-patch scenarios were recorded before this restore; their shared
system prompt sidecar still shows the old read sentence.
2026-11-02 16:42:47 +08:00
sjw1231 822229c588 fix(tool-fs): update apply_patch tool description
Describe what the patch tool does and the patch text it expects, and
refresh the tool catalog and recorded schema snapshots to match.

Refs #7696
2026-11-02 15:26:37 +08:00
Dudu-0223 ff563d07e0 feat(subagent): name subagents in their creator catalog
Models coordinate subagents through send_message, interrupt_agent, list_agents,
and Team task owners, but every one of those addressed a member by a
36-character Session id. Each new child now has a name, unique among its
creator's children, recorded as a field of the subagent/catalog entry the
creator already appends (payload version 3).

The Session id stays the only execution identity for routing, authorization,
persistence, task ownership, and navigation. A name is a model-facing
reference: 1-32 lower-case letters, digits, or hyphens starting with a letter,
never `lead`, so a UUID never parses as one. Omission selects the first free
`agent-N`; an explicit duplicate or invalid name fails before startup. The
creator reserves the name synchronously at admission and releases it when the
catalog entry commits or startup fails, so no second write, rollback protocol,
or optional naming service exists.

The root is `lead`, and a descendant's full name is its path from the root
without a `lead/` prefix, such as `reviewer/checker`. Sibling-unique segments
make full names tree-unique without a root registry, and cold resume rebuilds
a member's path from the catalogs along its creation chain. A member below an
unnamed historical creator keeps its Session id as its name. resolveName()
reads only the catalogs along the path and grants no authority.

Tools, runtime-authored text, and the Web client consume names: delegation
takes an optional name and returns { kind, subagentId, name }; send_message and
interrupt_agent take `agent`; list_agents rows carry name, status, and
description; Team task results report the owner's name while storage keeps
ownerId; identity runtime context, message framing, and settlement notices use
full names. Recorded sessions, CLI and Web expectations, the Python runtime
snapshots, bilingual docs, an upgrade guide, and an Agent Note follow.
2026-11-02 14:58:32 +08:00
Dudu-0223andimccyu 0487699ae4 feat(subagent): share ID-based identity and collaboration guidance (#7519)
* feat(subagent): share identity and collaboration guidance

* fix(subagent): complete shared guidance snapshot expectations

* fix(subagent): match task descriptions to the user language

* refactor(subagent): retain shared guidance with Session ids

* test(subagent): recognize parent ids in shared identity

* Checkpoint ID collaboration guidance generated expectations

* Preserve historical Web schema sidecars

* fix(test-support): preserve job references during snapshot refresh

* test(subagent): refresh ID-only collaboration guidance outputs

* fix(subagent): rely on list_agents tool description

* fix(subagent): keep collaboration guidance consistent with live identity

* test(team): parse parent identity from shared guidance

* test(subagent): address parent from shared identity context

* test(subagent): refresh compiled Web guidance snapshots

* test(subagent): refresh browser collaboration context fixtures

* docs(subagent): attribute required identity to standard controls

* fix(subagent): clarify collaboration identity prompts

* test(subagent): remove duplicate PowerShell collaboration context

* refactor(subagent): match unavailable-root errors by name and code

---------

Co-authored-by: imccyu <[email protected]>
2026-10-31 22:38:07 +08:00
Dudu-0223 63ac6b46a6 test: refresh merged subagent tool schemas 2026-10-30 16:02:36 +08:00
sjw1231 cc2718bede fix(fs): address review of the shipped mutation defaults
A new Agent Note owns the shipped defaults and the original apply_patch note
keeps its opt-in decision with a supersession link; the policy README, the
tool-catalog note, the filesystem page, and the fs group README state the
current defaults without evaluation figures. The PTC overlay sets
`applyPatch: false` so the SDK and Desktop PTC tool sets match without a
parity exception; the fixed-budget compaction scenarios pin the tool set in
their live patches too; the apply-patch composition drops its no-op override.
The upgrade guide scopes the change to base-backed profiles, the README
refusal example is one every composition can produce, the integration suite
names its deployments by guard, and the keyless SDK smoke edits a file the
model never read through the shipped profile.
2026-10-29 21:19:14 +08:00
sjw1231 14d3e53c19 feat(fs): ship apply_patch on and the observation policy off
Evaluation on the shipped compositions decided the defaults: the `dsh-base`
`tool-fs` row and the Web `standard` preset enable apply_patch, keeping the
SDK and Desktop `standard` tool sets equal (the SDK PTC overlay keeps the
base row while the Desktop PTC preset lists its own, which the parity
exceptions record); the `fs-observation-policy` row ships `disabled: true`,
so write, edit, and apply_patch mutate unconditionally in every profile, and
the package is classified as an optional product package.

The write and edit guidance stops citing the policy as the default and the
apply_patch guidance asks for enough hunk context instead of a prior read.
The two recordings that capture a policy refusal re-enable the row through a
per-scenario runtime.cordis.yml; the fixed-budget compaction scenarios pin
`applyPatch: false` so their recorded thresholds hold; every other sidecar
and recorded header follows. An upgrade guide records the default change.
2026-10-29 18:36:52 +08:00
sjw1231 ea4837abb1 feat(fs): keep apply_patch opt-in and the observation policy mounted
This pull request lands the tool; switching the shipped defaults is a
separate pull request stacked on it. Every shipped `tool-fs` row states
`applyPatch: false`, the `fs-observation-policy` row stays mounted, the
write, edit, and apply_patch guidance names the policy, and every recorded
expectation outside the apply-patch scenarios returns to the master
baseline. The upgrade guide, the parity exception, and the optional
classification of the policy package move to the stacked change.
2026-10-29 18:36:31 +08:00
sjw1231 fe5450b152 feat(fs): ship without the observation policy and enable apply_patch in the Web standard preset
Evaluation on the shipped Web composition decided the defaults: the
`fs-observation-policy` row in dsh-base ships `disabled: true`, so write,
edit, and apply_patch mutate unconditionally in every profile; the Web
`standard` preset lists apply_patch; and the tool-fs `applyPatchGuard` switch
is gone, apply_patch always consulting the fs/* intent waterfalls like the
other mutating tools.

The write and edit guidance stops citing the policy as the default, and the
apply_patch guidance asks for enough hunk context instead of a prior read.
The two recordings that capture a policy refusal (fs-policy-reject and
apply-patch-rejected) re-enable the row through a per-scenario
runtime.cordis.yml; every other sidecar follows the prompt change, and the
Web standard-preset recordings gain the apply_patch schema. An upgrade guide
records the default change.
2026-10-29 16:38:42 +08:00
sjw1231 5711d3aae7 docs(tool-fs): state apply_patch's whole-patch preflight in the model guidance
The guidance described only the commit phase ("stop at the first failure,
not rolled back"), so a model inferred that an invalid hunk would leave
earlier operations applied. The sentence now says the whole patch is checked
first and a rejected patch changes nothing before describing commit order.
2026-10-28 15:37:36 +08:00
sjw1231 1f1a9cda72 feat(tool-fs): add applyPatchGuard to run apply_patch without observation guards
`tool-fs.applyPatchGuard` selects the mutation guards of apply_patch:
`observation` consults the mounted observation policy per operation like
write and edit; `none`, the default, applies hunks unconditionally on
context match alone as the Codex tool does, keeps publishing fs/observed,
and replaces the read-first guidance with a context-matching reminder.

In 2,226 recorded DS-Bench and DeepSWE rollouts with the policy on, read-first
and stale-version refusals were 29–41% of apply_patch failures although hunk
context matching already rejects a divergent file. The recorded apply-patch
scenarios pin `observation`, under which they were captured.
2026-10-28 15:26:21 +08:00
sjw1231 64d04a4916 feat(tool-fs): name the apply_patch argument input as Codex does
Across 2,226 recorded DS-Bench and DeepSWE rollouts, 14% of apply_patch
calls arrived as {"input": ...} and were rejected for the missing `patch`
property: models trained on the Codex tool send its parameter name
unprompted. The schema now uses `input`; the executor request keeps its
internal `patch` field. Recorded apply-patch sessions, the schema sidecar,
and the catalogs follow.
2026-10-27 22:26:39 +08:00
sjw1231 d37b2ba52b Merge branch 'master' into feat/apply-patch
Resolves the headless snapshot runner (keeps both the readonly-patch-directory
and cua-script-guidance workspace setups), regenerates the client slot and tool
catalogs with their pairing record, adds the apply_patch tool title to the five
locales master introduced, and refreshes the apply-patch system-prompt pin for
master's subagent guidance change.
2026-10-26 18:01:46 +08:00
sjw1231 81b0bcf3ef feat(tool-fs): make apply_patch an opt-in switch, default off
`tool-fs.applyPatch` now defaults to false instead of deriving from
`write && edit`, so every shipped composition keeps the model request it
had before the tool existed. The `tool-fs` rows that enable `edit`
(dsh-base and the Web standard, ptc, and cordis presets) state
`applyPatch: false` explicitly; changing that one line adds the tool.

The apply-patch snapshot composition enables the switch, the tool catalog
mounts tool-fs with it on so the schema stays documented, and every other
recorded expectation returns to the master baseline.
2026-10-26 17:12:36 +08:00
sjw1231 cdd7c76f2a Merge branch 'master' into feat/apply-patch
# Conflicts:
#	apps/cli/tests/profiles/headless/tests/expected/subagent-inheritance/parent.expected.jsonl
#	docs/config-catalog.md
#	docs/config-catalog.zh.md
#	packages/fs/tool-fs/README.i18n.yaml
#	scripts/verify-package-readme-model-experience.ts
#	snapshots/sdk/dynamic-tool-updates/notifications.expected.jsonl
#	snapshots/sdk/dynamic-tool-updates/session.v4.jsonl
#	snapshots/sdk/tool-error-details/notifications.expected.jsonl
#	snapshots/sdk/tool-error-details/session.v4.jsonl
#	snapshots/session/advanced-toolchain/session.v4.jsonl
#	snapshots/session/dynamic-tool-prompt-updates/session.v4.jsonl
#	snapshots/session/dynamic-tool-updates/session.v4.jsonl
#	snapshots/session/pi-ai-conversation-updates/session.v4.jsonl
#	snapshots/session/plugin-manager-mcp/session.v4.jsonl
#	snapshots/web/collapse-timing/next-message.expected.md
#	snapshots/web/collapse-timing/verbose.expected.md
2026-10-22 16:23:14 +08:00
sjw1231 8f798da97d fix(tool-fs): leave sandbox decisions to providers and bound patch diffs
Drop the preflight sandbox fence. It derived writable roots in the local
process, so under the SSH provider it threw on remote additionalWritableRoots
and compared remote targets against local temp roots; write and edit never
fenced in the tool layer. The provider now refuses a denied operation at
commit, and a commit whose first operation failed without any change rejects
like a preflight refusal so the sandbox escalation path keeps its error code.

Classify FS_ABORTED as unknown: a cancelled remote request rejects while the
provider may still be writing, so no stat comparison can settle it. Document
that a cancelled call receives the registry's abort result like every tool.

Store no diff for creations, matching write, and bound the move source basis
by diffBasisMaxBytes so persisted metadata never holds whole files;
FsMoveOutcome.sourceBefore becomes nullable. Omit deletions from the call
preview and from result diffs without a basis instead of rendering them as
empty creations, and count distinct files in the preview title.

Reject a NUL byte beyond the read sampling window in preflight, name targets
by display path in atomic-write diagnostics, align the apply_patch guidance
with the edit wording and mention delete and move, read the stopped flag and
applied diffs tool-agnostically in the client models, skip the permission-bit
snapshot under root, list fs/delete-intent in the tool catalog, and refresh
the fs-patch description, group README, fs-sandbox summary, and ui-tool README.
2026-10-21 17:21:01 +08:00
sjw1231 8834433444 Merge branch 'master' into feat/apply-patch 2026-10-20 20:25:24 +08:00
sjw1231 10fd0f806c feat(fs): enable apply_patch by default 2026-10-16 16:40:58 +08:00
sjw1231 42958baa8c test(fs): record apply_patch success rejection and partial execution 2026-10-16 12:03:35 +08:00
sjw1231 7ab485156e fix(tool-fs): report move destinations and remove unrecorded scenes 2026-10-15 15:29:21 +08:00
sjw1231 f7dae57fda test(snapshot): prepare apply-patch scenes and document tool activation 2026-10-15 03:51:09 +08:00