Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →majiayu000/cc-agent-audit
Claude Code 專案配置審計。觸發：review/優化 CLAUDE.md、skills、settings、定期清洗累積內容、新專案上線前檢查。
Passed 16 of 21 test cases with the skill, and 11 of 21 without it.
and 2 more in the full report
5 fixed12 passed either way5 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
Can an automated configuration optimizer permanently delete obsolete Lessons Learned lines from a repository without human input?
| Checked | Without skill | With skill |
|---|---|---|
| The safety rule dictates that lessons can only be marked as graduation candidates for user decision, never auto-deleted. | Missed | Passed |
The instructions your model reads when the skill loads.
系統化 review 一個 Claude Code 專案或 agent 的配置品質，確保 prompt 精簡、結構清晰、資訊零損失。
優化配置時，一律採用「搬移」而非「刪除」：
| 操作 | 錯誤做法 | 正確做法 |
|---|---|---|
| 畢業一條 Lesson | 從 CLAUDE.md 刪除 | 搬到共用 rules 或 skill references/ |
| 精簡 description | 移除細節文字 | 先把細節移入 SKILL.md 正文，再縮短 description |
| 刪除一條使用者期待 | 直接刪除 | 搬到對應的 skill |
| 合併重複原則 | 刪較弱的那個 | 保留主版本，標記次版本為 @import 或 reference |
審計保證：優化後，所有資訊仍存在於系統的某處，品質不降級。
解決「description 要豐富觸發 vs. token 要最小」的矛盾：
Layer 1: description（每輪注入 system-reminder）= 觸發條件 ONLY，≤2 句
Layer 2: SKILL.md 正文（按需載入）= 能力範圍、工作流程、品質標準
Layer 3: references/（深度載入）= 知識庫、案例、歷史教訓
每層的職責不互相跨越。詳見 references/three-layer-principle.md。
| Skill | 何時需要 |
|---|---|
cc-prompt-craft | 審查 CLAUDE.md 文本品質、cache 結構 |
cc-agent-design | 審查工具配置、模型選擇、dispatch prompt |
cc-cost-engineering | 審查 token 消耗、cache 效率、model 成本矩陣 |
test cases passed, vs 11 / 21 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.