agent-design
Skill4 starsChanged 4 months agoMITView on GitHub
Use when creating a new custom agent/subagent definition (.claude/agents/*.md), auditing or rewriting existing agent definitions, deciding whether a task deserves a dedicated agent vs general-purpose with a good prompt, choosing an agent's tools whitelist or model tier, or when delegated agents keep going off-scope, dumping full transcripts back, or overlapping each other's roles.
13 other projects use a version of this.
Adopted by
Most recently taken up by
- mariomile/builder-ostheir own version, first seen Nov 1, 2026
- xiaolai/cc-suitetheir own version, first seen Oct 19, 2026
- bingyue/fde-skillstheir own version, first seen Oct 13, 2026
- The-Utopia-Studio/skillstheir own version, first seen Oct 13, 2026
- nimadorostkar/Claude-Skills-collectiontheir own version, first seen Oct 13, 2026
- BULDEE/ai-craftsman-superpowerstheir own version, first seen Sep 21, 2026
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- wibus-wee/cradle-app★ 75same name
- nahid-sparktales/agent-dispatcher★ 59same name
- xiaolai/cc-suite★ 56same name
- BULDEE/ai-craftsman-superpowers★ 35same name
- nimadorostkar/Claude-Skills-collection★ 33same name
- The-Utopia-Studio/skills★ 6same name
- bingyue/fde-skills★ 5same name
- Minions-Land/Soup_Drinking★ 4same name
- freezingfunky/ashwin-skills★ 2same name
- gmanch94/claude-code-aidlc-template★ 2same name
3 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 repository12 other files it gives its agents.
- android-dev-playbookSkill
- backend-server-playbookSkill
- dispatching-subagentsSkill
- feature-flowSkill
- flutter-dev-playbookSkill
- frontend-web-playbookSkill
- go-concurrency-playbookSkill
- harness-writingSkill
- ios-dev-playbookSkill
- mobile-ui-playbookSkill
- rag-playbookSkill
- thinking-kernelSkill
The file
Files
🌐 English version · 繁體中文(正本 / canonical)
Agent 設計判準(agent-design)
定位:dispatching-subagents 管怎麼用 agent,本 skill 管怎麼造 agent。
核心認知:agent 定義是給弱模型的合約,不是履歷。「你是 15 年經驗的專家」一句就夠,
真正決定行為的是:負面邊界、工具白名單、輸出契約。
核心原則
- 預設不建新 agent:general-purpose + 派工三件套 + 對應 playbook skill 已覆蓋八成需求。
- 人設 ≤2 行,邊界才是本體:能力清單(「你會 A、B、C」)對行為幾乎無約束力;「你不做 X、做到 Y 為止」才有。
- 工具白名單是機器強制的邊界:建議型 agent 給 All tools=把「希望它不亂改」放在最弱的紀律層。
- 沒有輸出契約的 agent 會把整個過程倒回主對話——每個 agent 必有 Output Format。
開工分流
| 情境 | 路徑 | 先讀 |
| 想建新 agent | 先過「該不該存在」三問,過了才寫 | references/design-criteria.md §1 |
| 審計/整修既有 agent | 逐個跑四件套檢查表 | references/design-criteria.md §2 |
| 選 tools / model | 查最小權限表與 tier 對應 | references/design-criteria.md §3–4 |
| 新 agent 上線前 | 逐條跑必查清單 | references/release-checklist.md |
「該不該存在」三問(全部否 → 不建,用 general-purpose)
- 這個 agent 的系統 prompt 是否帶來穩定的行為差異?(如審查者的對抗立場、驗證者的不信任預設——這種「立場」值得固化)
- 是否需要工具白名單限制它?(唯讀審查、禁改檔的建議者——需要機器強制就值得建)
- 是否高頻重複委派同型任務,每次手寫 prompt 成本高?
- ❌ 「Flutter 專家 agent」——知識型人設。知識已在 playbook skill 裡,general-purpose 被指示「先讀 flutter-dev-playbook」效果相同,多一個 agent 只是多一個選擇困難。
- ✅ 「code-reviewer agent」——對抗立場(不信任實作者)+ 唯讀白名單 + 高頻使用,三問全過。
紅線(絕對禁止)
- 禁止建議/審查型 agent 拿到 Edit/Write——它的產出是意見,能改檔案的意見遲早變成未經授權的改動。
- 禁止沒有 Output Format 的 agent 定義——回報格式缺失的代價由每一次委派支付。
- 禁止只有能力清單、零負面邊界的定義——那不是 agent,是慫恿。
- 測試類 agent 的定義裡必須寫死「禁止為了變綠改鬆斷言/skip」——這條紅線不寫進合約,弱模型在 CI 壓力下必犯。
- 禁止人設段落超過 2 行——多出來的字是 token 稅,不是能力。
失敗訊號(該回頭,不是重試)
| 徵兆 | 多半是 | 修法 |
| 兩個 agent 的 description 對同一任務都成立 | 定位踩線 | 合併,或寫互鎖條款(「X 移交給 Y」) |
| 委派結果總是倒回全過程流水帳 | 輸出契約缺失 | 補 Output Format + 回報合約 |
| agent 清單 >10 個、每次選擇都要想 | 建太多了 | 跑三問淘汰,知識型的降級為 playbook skill |
| agent 老是做超出請求的事 | 零負面邊界 | 補「不做什麼」≥3 條 |
references 索引
references/design-criteria.md— 三問詳解、四件套檢查表、tools 最小權限表、model tier、正反例對照。造/審 agent 前讀。references/release-checklist.md— 新 agent 上線前逐條必查。references/test-scenarios.md— 判準測驗集。不給執行中的模型讀。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

