Repository navigation
chore(agents): 追蹤 .claude 自製 agent 資產 - #201
Merged
Merged
Conversation
…s/settings） .gitignore 原本以 .claude/* 排除全部，僅白名單 ship-item 與 gitnexus-blast-radius，導致大量自製資產（16 個 workflows、7 個 commands、3 個 agents、spec-to-done 與 omniverse 系列 skills、 hooks 設定）只存在本機，無備援。 - 擴充 .gitignore 白名單：workflows/、commands/、agents/、 settings.json 全追蹤；skills 增列 spec-to-done 與 omniverse-* 三項 - 第三方安裝的 skills（matt-pocock 系列、gitnexus 系列、generated/） 維持排除，重裝即可還原 - settings.local.json、worktrees/ 維持排除 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
|
Important Review skippedToo many files! This PR contains 300 files, which is 150 over the limit of 150. To get a review, narrow the scope: ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (300)
You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4bbb08f20e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "hooks": [ | ||
| { | ||
| "type": "command", | ||
| "command": "powershell.exe -NoProfile -ExecutionPolicy Bypass -File \"C:/Repos/active/iot/AI-BIM-governance/scripts/claude-commit-guard.ps1\"" |
There was a problem hiding this comment.
In Linux/macOS workspaces or any clone that is not at C:/Repos/active/iot/AI-BIM-governance, this repo-level Claude setting runs before every Bash tool use but invokes powershell.exe with a hard-coded Windows-only script path. That makes the commit guard fail or become ineffective for other contributors/agents as soon as they pull this tracked .claude/settings.json; use a repo-relative/cross-platform hook or gate this to the Windows checkout where the script exists.
Useful? React with 👍 / 👎.
| // 誠實鐵律在程式層 enforce:宣稱 slice 綠但沒截圖或沒 summary JSON 落檔 = 不可信 | ||
| // (summaryJson 是 P6 前置 PR 10 列表的資料來源,缺它通道會斷在 P6) | ||
| // detail 帶 probe 結果(含 stack_down / port 占用註記),SKILL P4 gate 靠它分流 golden-path 重啟 | ||
| if (!ev.verticalSliceOk || ev.notObserved.length > 0 || ev.screenshots.length === 0 || !ev.summaryJson) { |
There was a problem hiding this comment.
When the evidence agent reports screenshot or summary paths that were not actually created or were left in a disposable worktree, this gate still returns ok: true because it only checks that screenshots.length is nonzero and summaryJson is truthy. That can let a user-facing change pass the browser-evidence gate with no reviewable artifact on disk, despite the comment saying the workflow enforces persisted screenshots and summary JSON; validate the returned paths exist in artifacts/e2e before returning success.
Useful? React with 👍 / 👎.
|
|
||
| phase('Impact') | ||
|
|
||
| const allSymbols = [...new Set(plan.tasks.flatMap((t) => t.symbols || []))].filter((s) => !ACKED_CRITICAL.includes(s)) |
There was a problem hiding this comment.
If a plan reviewer forces the fixer to add or change task symbols, the impact prescan here still uses the original plan.tasks returned before the fixes were committed. In that scenario P1 can report a clean or lower-risk blast radius while the actual fixed plan now edits additional symbols, so the early CRITICAL/HIGH gate is evaluating stale data; re-read/re-parse the plan after successful fixes before building allSymbols.
Useful? React with 👍 / 👎.
| 2. Playwright(repo 事實標準)。依序跑,worktree 缺 node_modules 時先在 ${ROOT}/web-viewer-sample 跑 npm install(失敗才放棄);瀏覽器 cache 缺 chromium-* 時跑 npx playwright install chromium 補: | ||
| node --version | ||
| ls "${ROOT}/web-viewer-sample/node_modules/.bin/" | grep -i playwright | ||
| ls "$LOCALAPPDATA/ms-playwright" |
There was a problem hiding this comment.
On Linux/macOS agents, $LOCALAPPDATA is normally unset and Playwright stores browsers under a different cache path, so this probe can mark Playwright unavailable even when npx playwright test would work. That blocks every user-facing spec at P4 with no_browser_engine outside the author's Windows environment; prefer npx playwright --version/npx playwright install --dry-run or check Playwright's platform-specific cache instead of a Windows-only variable.
Useful? React with 👍 / 👎.
Comment on lines
+121
to
+122
| if (!plan) return { ok: false, held: 'plan_author_failed', planPath: PLAN_PATH } | ||
| log(`plan 完成:${plan.taskCount} tasks,committed=${plan.committed}`) |
There was a problem hiding this comment.
If the plan author writes the plan file but reports committed: false after a git failure, this workflow only logs the value and continues into review/implementation. That can leave the plan untracked or folded into a later task commit, breaking the per-phase commit anchors that resume and PR review rely on; treat committed !== true as a held plan-author failure before proceeding.
Useful? React with 👍 / 👎.
Contributor
PR Review Agent Summary
Blockers
Warnings
Validation Commands
Checks
Human Review Notes
|
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
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.
摘要
.gitignore原本以.claude/*排除全部，僅白名單 ship-item workflow 與 gitnexus-blast-radius skill，導致大量自製 agent 資產只存在本機、沒有任何備援。本 PR 把這些資產納入版本控制。變更內容
.claude/workflows/（16 個）：ship-item、std-* 系列、adversarial-verify 系列等全部追蹤.claude/commands/（7 個）：build / code-simplify / plan / review / ship / spec / test.claude/agents/（3 個 + README）：code-reviewer / security-auditor / test-engineer.claude/skills/增列自製項：spec-to-done、omniverse-cad-to-simready、omniverse-realtime-viewer、omniverse-usd-performance-tuning.claude/settings.json：commit guard 與 verify reminder hooks 設定維持排除
settings.local.json、worktrees/、launch.jsongenerated/），重裝即可還原驗證
hf_xxxx...佔位符，無真實機密🤖 Generated with Claude Code