Read-only 跨平台 code/architecture/repo/release/Skill 審計器。用於清理前盤點、重複/依賴/分層/長責任、sync、release/資安、生成資產、Skill Token、GitHub 自動更新與舊版淘汰稽核;也是 R&D 的 measurement provider。
Scanned 9/4/2026
Install to Claude Code
npx -y skills add gabrielmoreira/agent-skills-mirror --skill code-cleanup-helper --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Code Cleanup Helper?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/gabrielmoreira-code-cleanup-helper)More formats (shields.io, HTML) on the badges page.
---
name: code-cleanup-helper
description: Read-only 跨平台 code/architecture/repo/release/Skill 審計器。用於清理前盤點、重複/依賴/分層/長責任、sync、release/資安、生成資產、Skill Token、GitHub 自動更新與舊版淘汰稽核;也是 R&D 的 measurement provider。
---
# Code Cleanup Helper
## 永遠適用的 contract
1. Active private Skill 目錄是唯一權威。讀取前 `check_skill_revision.py capture` 一致性雙掃描,最終 `verify`;`UNSTABLE` 不混用,`STALE` 重讀並重跑受影響 gate。
2. Audit 永遠 read-only。純診斷先報告再等授權;已授權的 R&D/實作只取證,不重複詢問。不得自行做外部或破壞性 mutation。
3. `FAIL`=已證明阻擋、`REVIEW`=人工判斷、`NOT_CHECKED`=量尺/環境/證據缺口、`PASS`=僅已執行維度;未量測不得包裝為 PASS。
4. 外部檢查要保存 child executable identity、真實啟動、child exit 與 success marker;父 shell exit 0 不代表通過。
5. 解析目標絕對路徑;修後重跑同 audit,報告綁定 revision、config、route 與 JSON evidence。
6. Cleanup 是獨立唯讀 provider;R&D 可選 mode/config,但選模組不等於通過。修改、promotion、外部 mutation 與完成宣告歸 orchestrator。
7. 每次 audit 都分類下游 target 的更新覆蓋(managed/check-only/safe-auto-update/manual-only/no-origin);不由 URL/檔案存在升格,也不讓 Cleanup/R&D 自行更新。
## Progressive-disclosure 路由
只讀本次適用列;不確定、缺失或 budget overflow 時使用明示的 legacy fallback,不可靜默漏掉 critical rule。
| 任務 | 讀取 |
|---|---|
| duplicate/命名/架構/依賴/長函式 | `references/mode-a.md` |
| sync/release/link/drift/handoff | `references/mode-b.md` |
| config、例外、machine report、資產配對 | `references/config-and-report.md` |
| R&D baseline/promotion/freshness | `references/rd-integration.md` |
| capability/完整收尾 | `references/capability-obligations.md` |
| build/installer receipt | `references/build-receipt-audit.md` |
| security/secret/extracted payload | `references/security-and-release-hygiene.md` |
| model/prompt/Token claim | `references/model-context-contract-audit.md` + `references/topics/context-routing-and-memory.md` |
| 一般跨系統 journey | `references/topics/cross-system-core.md` |
| 專業媒體/Timeline/成片 | `references/topics/media-workstation.md` |
| desktop runtime/背景 job | `references/topics/desktop-runtime.md` |
| Codex/Claude/MCP 產品流程 | `references/topics/session-native-ai.md` |
| GitHub updater/回滾/淘汰舊版 | `references/topics/secure-self-update.md`;helper/installer/簽章再加 security+build |
跨類任務合併必要列;歧義時載入完整 legacy `references/cross-system-integration-audit.md` 並記錄原因/成本。修改本 Skill 讀 `references/maintenance.md`。
## 執行與報告
```powershell
python scripts/audit.py <target> --mode all --format json
```
僅 CI/fail-fast 加 `--strict`。普通 `REVIEW` 不阻擋;完整收尾由 R&D 收斂 review、freshness、obligations。未解析的非 Python graph 保持 `NOT_CHECKED`。輸出 3–10 個最高優先 finding,完整清單留 JSON;GitHub URL 不授權下載執行。
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!