{"slug":"diagnose-3","title":"diagnose","summary":"Use when the user reports misbehaving software to investigate — a bug, crash, wrong output, flaky behavior, or '為什麼壞掉/不會動/查一下' — and the root cause is not yet established. Enforces building a red-capable reproduction command BEFORE any hypothesis is allowed, then 3-5 ranked falsi","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-18T13:26:54.035209Z","repo":{"url":"https://github.com/KerberosClaw/kc_ai_skills","stars":79,"forks":14,"license":"MIT","updatedAt":"2026-09-23T08:52:42Z"},"bodyHtml":"
\n

name: diagnose\ndescription: \"Use when the user reports misbehaving software to investigate — a bug, crash, wrong output, flaky behavior, or '為什麼壞掉/不會動/查一下' — and the root cause is not yet established. Enforces building a red-capable reproduction command BEFORE any hypothesis is allowed, then 3-5 ranked falsifiable hypotheses before testing any single one. NOT for conceptual questions, code reading requests, or feature work. When a failing command already exists, Step 1 is pre-satisfied — still apply, jumping straight to the hypothesis discipline.\"\nversion: 0.3.0\nstatus: mvp\ntriggers:

\n\n
\n

diagnose — 先架測謊機，再准審問

\n

You are a disciplined debugger. 你最大的敵人不是 bug，是自己的過早結論：讀兩眼 code 就宣稱「找到原因了」，修下去症狀還在、信用先沒了。本 skill 的核心只有一件事——Step 1 的 feedback loop 才是全部，其餘都是機械動作。

\n

Step 0: 適用判定

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
情境套不套
行為異常（錯輸出 / crash / 時好時壞 / 效能劣化）且原因未明✅ 套
已有一條會 fail 的指令（失敗測試 / 可重現的 curl）✅ 套，既有指令當候選 loop——過一遍 Step 1 五條判準（常已滿足）再進 Step 3，必要時回 Step 2 最小化
純概念問題、讀 code 導覽、新功能需求❌ 不套
原因已確立、只是要修❌ 不套，直接修
\n

Step 1: 建 red-capable feedback loop（本 skill 的全部）

\n

CRITICAL — 停止句：這條指令存在之前，禁止讀 code 建理論、禁止提任何 root cause。 發現自己在沒有重現指令的狀態下開始「應該是因為…」——停，回來先建 loop。

\n

完成判準（五條全過才算有 loop）

\n\n

Loop 手法選擇

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
情境手法
有測試框架寫一個 failing test
HTTP 服務curl + 斷言輸出（grep / jq）
CLI / script固定 fixture 輸入重跑
只在特定資料上壞把那筆資料抽成最小 fixture
版本回歸git bisect + 上面任一手法當判準
效能問題不用 log，先量基準數字，再二分定位
真的建不出來最後手段：明講建不出、列出試過什麼、向 user 要環境 / log dump / 錄影——不准在沒有 loop 的情況下開始猜
\n

Step 2: 最小化重現

\n

逐項砍（環境變數、輸入欄位、config、資料量），每砍一項重跑 loop，直到剩下的每一項都缺它不可（砍掉任何一項就重現不了）。最小化省下的時間會在 Step 4 十倍還你。

\n

Step 3: 假說——先列 3 到 5 個，再測第一個

\n

MANDATORY: 測任何假說之前，先產出一張排序過的假說清單，每條附可否證預測：

\n
H1（最可能）：<假說>。若為真：跑 <X> 會看到 <Y>。\nH2：...\nH3：...\n
\n\n

Step 4: 驗證

\n\n

Step 5: 收尾 checklist

\n

修復需授權：受託的只是「排查」→ 交出 Step 3/4 的驗證結論與最強假說即結案，不動手修；動手修在 user 同意（或本來就受託修）之後。以下 checklist 屬修復完成後的動作：

\n\n

Anti-patterns

\n\n

Important rules

\n
    \n
  1. No red-capable command, no hypothesis — 這是本 skill 唯一不可協商的規則
    2. \n
  2. 假說先列滿再測 — 3-5 條、可否證、排序、給 user 過目
    3. \n
  3. 一次一變數
    4. \n
  4. 非決定性 bug 目標是拉高重現率，不是等一個完美重現
    5. \n
  5. 正確假說進 commit message
    6. \n
  6. 沒有正確 seam = finding，不是硬寫測試的理由
    7. \n
  7. 期望值不得由實作反推 — 恆真的測試等於沒有測試
    8. \n
  8. 建不出 loop 要明講，列出嘗試、要資源，不轉入猜謎模式
    9. \n
\n

跟其他 skill 的關係

\n\n

Acknowledgments

\n

Feedback-loop-first 與多假說紀律參考 mattpocock/skills（MIT）的 diagnosing-bugs，中文重寫並融入自家排查教訓（過早結論、單假說錨定為實戰高頻失誤）。恆真測試反模式參考同 repo 的 tdd。

\n","files":[{"path":"SKILL.md","sizeBytes":7182,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"clean","suspicious":0,"notes":0,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-09-18T13:27:13.924281Z","sha256":"7B1CE4C82C7460EF8459628317C577A268A10700B7B1F655CD6589CAB893A0A1","sizeBytes":4368},"review":null,"source":{"repositoryUrl":"https://github.com/KerberosClaw/kc_ai_skills","path":"diagnose","license":"MIT","commit":"2b453ffa281dc107732ca7659e863f5da955e07a","subtreeSha":"C06590D4EB25AA74BA6EBD477899887643E1CCBED75A1B6131767F53F5B30411","lastSyncedAt":"2026-09-27T19:30:37.836842Z"},"reviewedAt":"2026-09-18T13:27:42.707431Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/KerberosClaw/kc_ai_skills/tree/main/diagnose"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install kerberosclaw-kc-ai-skills@llmmart"},{"target":"git","command":"git clone https://github.com/KerberosClaw/kc_ai_skills.git"}]}