feat(u4-conduit-parity): --autonomy 起動宣言の導線を全面へ追記しパリティを回帰テストで固定する - #2532
Merged
Conversation
… every conduit The launch declaration landed in the engine with no entry point naming it: all eight harness surfaces, the help text, the README pair and the autonomy reference were silent, so a flag that works was undiscoverable. - name --autonomy on every harness entry point (six SKILL.md, two command files), with the real behaviour: none/semi apply at once, full prints the grant ceremony and stops - add --autonomy to the help text, the README pair and reference chapter 24 (both languages) - give semi its own decide-question procedure in stage-protocol, and revise AUTONOMY IS NEVER INFERRED so executing a recorded mode is not read as inference - add t492, a blocking parity guard whose conduit set is derived from the harness tree rather than listed
t492 observes the help text at the process boundary, so the repository-wide mechanism ratchet reclassifies it none->cli. Add it to EXPECTED_NONE_TO_CLI, which is the ratchet's declared spawner set.
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthrough
ChangesIntent autonomy conduit
Estimated code review effort: 3 (Moderate) | ~25 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
Comment |
Member
Author
|
@coderabbitai review |
|
…rio in the semi paragraph
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@docs/reference/24-intent-autonomy.ja.md`:
- Line 48: Specify the sh language on the startup-example code fence in
docs/reference/24-intent-autonomy.ja.md at lines 48-48 and
docs/reference/24-intent-autonomy.md at lines 50-50, updating both fences to
satisfy markdownlint MD040.
In `@packages/framework/core/amadeus-common/protocols/stage-protocol.md`:
- Line 137: Update the English and Japanese `semi` reference tables to include
non-boundary questions, matching the implementation contract of `SemiAuthority`,
`currentGrant = null`, saved policies, and the existing decision ladder when
`phase !== "phase-boundary"`. Keep phase-boundary and other milestone human
requirements unchanged, and add tests covering question handling, phase-boundary
behavior, and semi authorization.
In `@README.md`:
- Line 53: Update the “Intent autonomy” README entry in English and replace the
harness-specific `/amadeus --autonomy <none|semi|full>` example with
harness-neutral wording or a reference to the harness-specific launch commands.
Preserve the documented autonomy behavior, including launch-time application,
the human grant requirement for full autonomy, and recording unattended rulings.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: db990ab1-ca75-40ca-9a50-31a525c1378e
📒 Files selected for processing (16)
README.ja.mdREADME.mddocs/reference/24-intent-autonomy.ja.mddocs/reference/24-intent-autonomy.mdpackages/framework/core/amadeus-common/protocols/stage-protocol.mdpackages/framework/core/tools/amadeus-utility.tspackages/framework/harness/claude/skills/amadeus/SKILL.mdpackages/framework/harness/codex/skills/amadeus/SKILL.mdpackages/framework/harness/cursor/commands/amadeus.mdpackages/framework/harness/kimi/skills/amadeus/SKILL.mdpackages/framework/harness/kiro-ide/skills/amadeus/SKILL.mdpackages/framework/harness/kiro/skills/amadeus/SKILL.mdpackages/framework/harness/opencode/commands/amadeus.mdpackages/framework/harness/pi/skills/amadeus/SKILL.mdtests/integration/t-coverage-mechanism-ratchet.test.tstests/integration/t492-autonomy-conduit-parity.integration.test.ts
This was referenced Aug 8, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概要(Bolt: u4-conduit-parity / intent 260807-autonomy-reachability)
--autonomy起動宣言は u2(#2524)で実際に動くようになったが、conductor が読む面には記載がゼロだった(RE finding 8)。本 PR はその導線を全面へ追記し、パリティを blocking テストで固定する。harness/{claude,codex,kimi,kiro,kiro-ide,pi}/skills/amadeus/SKILL.md+harness/{cursor,opencode}/commands/amadeus.md)へ、フラグ列挙と説明ブロックを追記amadeus-utility.tsの help(:265 フラグ行 / :284 例)、README.md/README.ja.md、docs/reference/24-intent-autonomy.md/.ja.md(対訳は同一変更で同期)stage-protocol.md:137に semi の decide-question 操作手順を新設(fullと同じ手順を名指しし、実際の差分2点 — semi スコープの基準と--policies-file— だけを分離)。あわせてAUTONOMY IS NEVER INFERREDを「エンジンに記録された mode による自動裁定は推論ではない」旨と整合するよう改訂。同根棚卸しで claude だけでなく kimi/kiro/kiro-ide の同一行も是正tests/integration/t492-autonomy-conduit-parity.integration.test.tsを新設対象面の導出(件数のマジックナンバーなし)
導線集合はディスク走査で導出する —
readdirSyncでpackages/framework/harness/を列挙しisDirectory()で絞る(registry.ts/projections.tsは ignore list でなく構造的に落ちる)。各ハーネスは2つの入口形(skills/amadeus/SKILL.md/commands/amadeus.md)に解決する。新しいハーネスは存在するだけで契約に参加し、ファイル内に件数は一切現れない(count-free 原則)。導出自体も assert している:
entryPointsOf(harness).length !== 1は赤になる。これが無いと、入口が移動したハーネスが集合から静かに抜けても全 assertion が通ってしまう。docs 面(README 対 / reference-24 対)は列挙するディレクトリが無いため明示リストで、そのリストが canonical 定義になる。help text はソースを grep せず、
dist/claude/.claude/tools/amadeus-utility.ts helpを spawn してstatus === 0と stdout 内容で観測する。落ちる実証(builder / conductor が独立に1セットで実施)
conductor 側の独立再現:
packages/framework/harness/kimi/skills/amadeus/SKILL.mdから--autonomy4箇所を除去(残 0 を grep 確認)→ 赤がharness:kimiを名指しで発生(他4テストは green のまま)→git checkout f679e2864 -- <path>で復元 →git statusclean・残渣 grep 0 → 再実行 5 pass / 0 fail。builder 側は
opencode/commands/amadeus.mdで同型の1セットを実施済み。検証
bun run typecheck0 /bun run lint0 /complexity-gate --check0 /gen-coverage-registry --check0 /source-only:check0bash tests/run-tests.sh --ciexit 0(907 files・12181 assertions・0 fail)bun run build後の追跡ファイル不変申し送り
t492 は CLI を spawn するため mechanism ratchet が
none→cliと再分類しEXPECTED_NONE_TO_CLIへの登録を要求した(c7bf27864)。触ったテストだけを流しても発見できないゲートで、以後 spawn するテストを足す unit は同じ経路を通る。Refs #2378
Note
Low Risk
Documentation, harness prose, help text, and a parity integration test only; no changes to autonomy authorization logic in this diff.
Overview
Surfaces the existing
/amadeus --autonomy <none|semi|full>launch declaration everywhere conductors and users look, after the engine already supported it but harness entry points did not mention it.Harness parity: all eight harness canonical entry points (
skills/amadeus/SKILL.mdorcommands/amadeus.md) now list--autonomy, instruct passing it verbatim toamadeus-orchestrate.ts next, and spell out thatnone/semirecord via the canonical path whilefullstill requires the grant ceremony (not the flag). CLI help inamadeus-utility.tsadds the flag line and an example. README (EN/JA) anddocs/reference/24-intent-autonomygain a “declaring at launch” section with first-declaration-only rules andfull/revoke refusal behavior.Stage protocol: adds an explicit
semiquestion path using the samedecide-questionprocedure asfull, and tightens “AUTONOMY IS NEVER INFERRED” so engine-recordedsemi/fullrulings are not treated as inferred permission.Regression: new
t492-autonomy-conduit-parityintegration test derives every harness entry from disk, asserts--autonomyon each conduit plus doc surfaces and spawnedhelpoutput, and checks the semi paragraph exists instage-protocol.md.Reviewed by Cursor Bugbot for commit f679e28. Configure here.
Summary by CodeRabbit
none・semi・full）を指定できます。semiは限定的な自動処理、fullは人間による確認・許可を経て利用できます。semiとfullにおける質問処理や停止条件を明確化しました。