skill-author
Skill9 starsChanged 4 months agoMITView on GitHub
從零做新 skill 標準流程:派 worker 寫 script→codex 紅隊→補 SKILL.md。觸發:做一個 skill、製作 skill、新增 skill、skill-author
23 other projects use a version of this.
Adopted by
Most recently taken up by
- DylXBT/claude-orchestrator-skillstheir own version, first seen Nov 8, 2026
- himkt/cafleettheir own version, first seen Nov 8, 2026
- pankaj-mahaur/Agent-Daemontheir own version, first seen Nov 6, 2026
- educlopez/mallardtheir own version, first seen Nov 1, 2026
- mandasTeam/skill-authortheir own version, first seen Nov 1, 2026
- Lethe044/agentskill-kittheir own version, first seen Nov 1, 2026
Other versions23 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 23
- generative-computing/mellea★ 1.9ksame name
- aqm857886159/Nomi★ 550same name
- DylXBT/claude-orchestrator-skills★ 114same name
- robium-ai/robium★ 93same name
- himkt/cafleet★ 56same name
- Lethe044/agentskill-kit★ 48same name
- manusco/resonance★ 38same name
- Zhonghao1995/agentic-swmm-workflow★ 31same name
- felvieira/claude-skills-fv★ 27same name
- peter-tu-zynkr/zynkr-skill-builder★ 21same name
13 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- doc-to-skillSkill
- memory-cardsSkill
- skill-vettingSkill
The file
Files
skill-author
從零做新 skill 的 SOP。固化 2026-05-04 做 switch-channel-model skill 時走通的流程,避免下次重新摸索。
為什麼存在這個技能
skill-vetting 處理外來 skill、skill-optimizer 優化已存在 skill,但「從零自己做新 skill」沒有 SOP。實務上會踩坑:
- 派 glm-code 寫 script 被它自己的 permission 攔(
~/life-os/skills/寫不進去) - mini-agent 不在 PATH 沒設 invocation
- 沒走紅隊就交付 → 帶著 HIGH severity bug 上線
本 skill 把流程固化:派工優先、撞牆 fallback opus 自寫、紅隊強制。
6 步流程
0. 三層研究(搜尋優先於建造)
在動手之前,強制走三層:
Layer 1 — 現有 life-os 有沒有
ls ~/life-os/skills/找觸發詞重疊、功能相近的。有的話:是要擴展它,還是確實要新建?
Layer 2 — 社群有沒有現成方案 快速搜 gstack / GitHub / 社群(≤ 10 分鐘)。找到就讀,評估是否直接用或 fork。
Layer 3 — 他們的假設在我們脈絡裡哪裡錯了 看完 L1+L2 後,問:「我們的限制或需求跟他們的假設有什麼不同?」 這個 diff 就是新 skill 的存在理由。沒有 diff → 大概不需要新建。
三層都走完,再進 Step 1。
1. 釐清範圍
- 一句話定義:這 skill 解什麼問題?觸發詞是什麼?
- 對照 Step 0 確認觸發詞不重疊,且有 Layer 3 的 diff
- 估計需要:純 SKILL.md(純文件指引)/ + script / + reference 子目錄
2. 建目錄
mkdir -p ~/life-os/skills/<skill-name>/scripts # script 不需要就省略3. 主 session 寫 SKILL.md(worker 禁止寫,子代理紅線)
參考 skill-vetting 規範的 frontmatter 格式:
---
name: <skill-name>
description: >
動詞開頭一句功能描述。
觸發:觸發詞 1、觸發詞 2、使用場景。
不觸發:不適用場景(指向其他 skill 名稱)。
消歧:歧義關鍵字的判斷規則。
version: "1.0"
created: "YYYY-MM-DD"
---- 內容 ≤ 80 行,前 10 行必含「為什麼存在」段(記錄踩坑驅動,避免下次重蹈)
- 主體必須包含三段(缺任何一段視為未完成):
- Happy path:主流程,正常情況下怎麼跑
- Degraded path:工具失敗 / MCP 斷 / 檔案不存在時,怎麼降級但不中斷
- Guard(紅線):禁止做的事 + 為什麼(❌ ✅ 格式)
- 結尾必含「紅線」清單(❌ ❌ ✅ 格式)
4. 派 worker 寫 script(如果需要)
- 估 ≥ 10 步迴圈 → mini-agent;3-10 步 → glm-code
- 任務包格式(CLAUDE.md 規定):目標 / 輸入 / 產出 / 驗收 / 限制
- 撞牆 fallback:worker 兩次寫不進去(glm-code permission 攔最常見)→ opus 直接 Write,不要再 retry。理由見 memory
feedback_pragmatic_dispatch.md。
5. codex 紅隊(無差別必跑)
codex exec "<紅隊指示,列出要審的點>"紅隊指示模板:
- 列你期望它審的 N 個點(command injection / race / 邊界 / 誤觸發)
- 要求格式:finding 1: <描述> [severity: CRITICAL/HIGH/MEDIUM/LOW] + 整體結論 PASS/FAIL/需要修
- 找到 finding → 派回 worker(或 opus fallback)修 → 再跑一次紅隊 Round 2
- Round 2 PASS 才算收工
6. sanity check + 結案回報
bash -n <script>syntax check- 跑引數錯誤路徑(不擾動實際系統):no-args / bad arg → exit 3
- 真實流程驗收:對 no-op 目標跑一次(例如 model 切換成「跟現在相同」的 model 應該成功不變化)
- 結案回報必含:派工履歷、紅隊結論、規範違規透明化(如有)
Degraded Path(撞牆時的降級處理)
| 失敗點 | 降級方式 |
| Step 0 L2 搜尋無結果 / 超過 10 分鐘 | 記「無相關社群方案」,繼續 Step 1,不阻斷 |
| Step 3 worker 寫不進 skills/(permission 攔) | 不 retry,opus 直接 Write SKILL.md |
| Step 4 glm-code / mini-agent 兩次失敗 | fallback opus 自寫 script,記入結案報告 |
| Step 5 codex 無法執行 | 主 session 手動審閱紅隊點,逐一標記 severity,補 Round 2 |
| Step 6 真實驗收環境不可用 | 用 no-op 目標跑語法 + 邊界路徑,記「完整驗收待補」 |
所有降級都必須在結案回報中透明化,不能靜默帶著降級上線。
Guard(紅線)
- ❌ 跳過 Step 0 三層研究直接開始寫(「需求很明確」不是理由——Layer 3 diff 才是新建的正當性)
- ❌ 跳過紅隊直接交付(「這個應該沒問題」是腦補,不是驗證)
- ❌ Round 1 紅隊有 HIGH/CRITICAL 但只修一半就交付(HIGH 沒清掉等於沒過關)
- ❌ 用 worker 寫 SKILL.md(CLAUDE.local.md 子代理紅線:skills/ 核心層禁止 worker 寫入)
- ❌ skill description 沒寫「不觸發」「消歧」(觸發詞重疊會讓錯 skill 被呼叫)
- ❌ SKILL.md 缺少三段中任何一段(Happy path / Degraded path / Guard 三段缺一視為未完成品)
- ✅ 撞牆 fallback 用 opus 直寫,但紅隊不能省
不適用
- 改現有 skill 的小字 / typo → Edit 直接改
- 把外來 skill 引入 →
skill-vetting - 評估現有 skill 該怎麼改 →
skill-optimizer
Lineage
<!-- 自動聚合至 skills/CAPABILITY-MATRIX.md by gen-capability-matrix.py;勿刪 -->origin: "" status: active closeout_gist: ""
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

