feature-flow
Skill4 starsChanged 4 months agoMITView on GitHub
Use when a feature or refactor changes public contracts, schemas, shared integration points, cross-layer data flow, release behavior, or other high-risk architecture decisions.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- pankaj-mahaur/Agent-Daemontheir own version, first seen Nov 6, 2026
- awahids/noskilltheir own version, first seen Oct 15, 2026
- rcelebrone/DotAgentstheir own version, first seen Oct 14, 2026
- dung204/hono-starter-templatetheir own version, first seen Aug 23, 2026
- sameerhimati/claude-code-kittheir own version, first seen Aug 14, 2026
- mateonunez/skillstheir own version, first seen Jul 21, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- rcelebrone/DotAgents★ 15same name
- mateonunez/skills★ 8same name
- awahids/noskill★ 1same name
- dung204/hono-starter-template★ 1same name
- sameerhimati/claude-code-kit★ 1same name
- pankaj-mahaur/Agent-Daemon★ 0same name
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.
- agent-designSkill
- android-dev-playbookSkill
- backend-server-playbookSkill
- dispatching-subagentsSkill
- 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)
feature-flow — Harness 的 high-risk-feature profile
核心原則
本 skill 提供高風險功能的 spec/design/implementation 技巧與 artifact template;Harness 存在時,風險路由、狀態轉移、證據與契約生命週期一律由 Harness 決定。本 skill 不建立第二套狀態機。
若專案沒有 Harness,仍沿用下列 profile,但先建立 risk assessment 與單一產出根目錄 docs/plans/<slug>/。
何時用
適用:public API/schema/共用整合點、跨層資料流、migration、release、auth/付款/正式資料,或需求歧義與回滾成本都高的功能。
不適用:只因檔案多;低風險機械變更走 quick/standard。若 design 寫不出實質取捨、也沒有絕對升級條件,回 risk router 降級。
Profile 狀態
1. Discovery / Risk
產出 plan.md 的 Risk Assessment:命中的 high-risk 絕對條件、quick 條件為何不成立、命中契約、既有可複製模式、外部副作用與驗證限制;不做數字加總。
檔案數只記錄爆炸半徑,不直接決定 profile。
2. Spec
產出 spec.md:Why、in/out scope、GIVEN-WHEN-THEN AC、假設、非目標與 rollback 條件。
只有昂貴歧義、產品/scope 取捨、public contract、不可逆或外部副作用需要使用者確認。命名與可逆內部選擇自行決定並記假設。
UI/UX 歧義必附 mockup/行為走查/截圖標註,讓錯誤理解在實作前暴露。
3. Design / Plan
產出 plan.md:架構與資料流、Key Decisions、介面、migration/rollback、phase、每 phase 的驗收,以及每個 AC 預定取得的 evidence。
每個 phase 必須可獨立驗證;若一個 phase 同時跨多個風險面,先拆分。既有同性質 pattern 優先於通用設計。
若有兩個以上工作包,plan.md 必須加入 Execution Graph,明列 dependency、writable scope、validation、可平行對象、handoff 與 integration owner。平行實作前先固定 shared contract 與 ownership;寫入重疊或無法安全切分者改為序列執行並記原因。計畫仍須可由單一 agent 依 dependency 執行。
4. Implement
可行時,每 phase 先取得會因缺少新行為而失敗的 test/schema check/模擬執行/diff assertion,再做最小實作;不可行時記錄原因與替代的 before/after evidence。機械檢查紅燈不得轉移。
實作推翻設計假設時,在 plan.md 加「偏離記錄」與新 evidence plan;不要偷偷改 scope,也不要為了文件一致性維持錯誤設計。
5. Verify / Review
evidence.md 逐條記 claim→evidence。Executable verifier 親跑 AC、build、test 與 baseline;semantic reviewer 審需求、契約、架構與 scope;auth/資料/release/不可逆面再加 adversarial challenge。
相同模型、相同 prompt、相同 evidence path 不算獨立冗餘。reviewer 從 diff、AC、evidence 與命中契約開始;若能指出明確風險理由,可以擴張唯讀探索並記錄範圍。無 fresh reviewer 時記錄「獨立性受限」,用實跑與 fresh read-back 補足。
6. Close
只有所有必要 claim 有證據、無新增 baseline failure、diff scope audit 通過,才能完成。review.md 記 residual risks、rollout/rollback 與未做項目。
只有使用者糾正、同類問題再次發生、incident/escaped defect、未文件化的重複限制,或新證據推翻既有契約時才建立 contract-candidates.md;沒有觸發事件就不建立空檔。一次觀察只能是 candidate,需再次發生、使用者裁決或事故證據才升 active。
Evidence 最低要求
- 每個 AC:至少一個 test/command/read-back/manual evidence,含來源與結果。
- 回歸:與開工基線比較,不只寫「全綠」。
- scope:列出變動檔與其對應 task;範圍外既有 bug 另記、不順手修。
- release/實機限定行為:實際環境證據;debug/模擬器不能代替。
失敗訊號
- 同一設計假設兩輪仍失敗 → 回 Design,不繼續加例外。
- AC 無法驗證或無法達成 → 回 Spec 調整 scope,需使用者裁決。
- 一個 phase 同時打開多個無關風險面 → 回 Plan 拆分。
- 平行工作包的 shared contract 未固定或 writable scope 重疊 → 回 Plan 建立前置工作包或改為序列執行。
- reviewer 只有意見沒有 evidence gap/失敗情境 → 降為建議,不阻擋轉移。
- 流程同時寫入
specs/changes/與docs/plans/→ 停止,選定docs/plans/<slug>/為 canonical。
紅線
- 未經明確授權 push、merge、deploy、刪除或執行不可逆操作。
- 以 reviewer PASS、CI 綠燈或 checklist 勾選代替功能證據。
- 因為「3+ 檔案」就自動使用本 profile。
- 把一次性偏好直接升成 active 契約。
references 索引
| 檔案 | 何時讀 |
references/artifact-templates.md | 建立實際適用的 spec.md/plan.md/evidence.md/review.md/contract-candidates.md |
references/lessons.md | 路由拿不準、review 獨立性、設計被推翻或流程成本過高時 |
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.

