dev-plan
Skill1 starsChanged yesterdayMITView on GitHub
需求分流、编写与审核四层工程文档(TDD、PRD、SPEC、Todo)、门型评估与规划内审自净化。 触发词:需求规划、写方案、写设计文档、写 PRD、写 SPEC、写 Todo、拆任务、 需求分流、立项、先出方案、文档自检、plan feature、write spec、triage task。
11 other projects use a version of this.
Adopted by
Most recently taken up by
- AIBiz-Automatyzacje/claude-code-startertheir own version, first seen Nov 14, 2026
- gaoyu06/harnesstheir own version, first seen Nov 8, 2026
- AIBiz-Automatyzacje/claude-crontheir own version, first seen Nov 2, 2026
- TchiRubick/skillstheir own version, first seen Oct 20, 2026
- XGENT-ai/skillstheir own version, first seen Oct 13, 2026
- egoxx1/unified-master-harnesstheir own version, first seen Oct 11, 2026
Used with
Projects with it also use dev-brainstorm.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
phdengSkilldev-brainstorm
结构化开发头脑风暴助手。用于:(1) 新项目/功能的需求分析与方案设计,(2) 技术选型讨论与权衡分析,(3) 架构设计与系统拆解,(4) 问题解决与调试思路梳理,(5) 代码重构与优化方向探索。当用户说"帮我想想"、"头脑风暴"、"怎么设计"、"技术方案"、"架构讨论"时触发。
Also in 3 projects that use this
Other versions11 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 11
- EvolutionAPI/evo-nexus★ 545same name
- evolution-foundation/evo-nexus★ 535same name
- AIBiz-Automatyzacje/claude-code-starter★ 94same name
- okazuki58/agent-skills★ 7same name
- gaoyu06/harness★ 3same name
- calneymgp/solodev★ 2same name
- AIBiz-Automatyzacje/claude-cron★ 1same name
- egoxx1/unified-master-harness★ 1same name
- ruccess/skills★ 1same name
- TchiRubick/skills★ 1same name
1 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 repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- dev-exploreSkill
- dev-flowSkill
- dev-governSkill
- dev-reviewSkill
- project-forgeSkill
The file
Files
dev-plan:需求 → 澄清 → 设计 与独立 Critique 网关
负责 8 阶段研发闭环中的需求整合、多轮互动澄清、分层设计定案与独立 Critique。只写经过勘察核实的真相,用恰如其分、不臃肿的文档冻结 C1..Cn 验收标准。
前置侦察、多链接协同与多轮互动澄清
- 先
dev-explore后动笔:起草任何规划前必须先通过repo-map.mjs查清真实文件、调用链、CI 命令与reusableModules,禁止凭记忆编造路径或重复造轮子;若收到多个 Issue/PR URL,先用flow.mjs inspect [URL...]联查互指关系与依赖拓扑序再统一设计。 - PRD 双触发入口:
- 用户明确提出写 PRD;
- 经
dev-explore勘察发现跨多模块、流程不清或业务痛点较复杂时,主动向用户建议先建总控 PRD(支持 1 份总控 PRD 拆分为 N 份切片SPEC + Todo)。
- 多轮互动澄清直至「无待拍板问题」:
- 遇到范围边界、接口取舍、异常降级策略或兼容性不明确时,主动停下来向用户提问澄清(可多轮互动),把每轮拍板结果直接补充完善进 PRD/SPEC/Todo 的对应章节与
C1..Cn; - 硬门槛:文档中「待拍板事项 / 未决问题」全部消解且路径/命令核实后,才可进入
audit与独立 Critique(严禁把未拍板问号留给实现或审查)。
- 遇到范围边界、接口取舍、异常降级策略或兼容性不明确时,主动停下来向用户提问澄清(可多轮互动),把每轮拍板结果直接补充完善进 PRD/SPEC/Todo 的对应章节与
- 尊重仓库自有模板与短名图表:
doc-plan.mjs init自动识别并套用仓库既有的docs/templates/*.md/.github/*-template.md,同时注入 YAML Frontmatter、元信息概览表与 SPEC 短名 Mermaid(矩形/分组单行<= 16、菱形/边<= 12单位,禁节点内 HTML 标签)。 - 说人话四要素 + 三不写:
- 四要素:真实痛点与目标、本期明确不做(防膨胀)、可运行/可断言的
C1..Cn、门型(two-way/one-way)与回滚路径。 - 三不写:不写空泛废话(如「提高质量」「确保稳定」「优化体验」)、不写假路径、不在文档里贴大段实现代码。
- 四要素:真实痛点与目标、本期明确不做(防膨胀)、可运行/可断言的
全场景五档分流(轻重恰当,绝不臃肿)
| 场景 / 档位 | 文档产物 | 规划内审与执行硬门禁 |
| 外部 PR 审查 | 0 文档 | 直接移交 dev-flow + dev-review 只读隔离扫描 |
| Case 1(问答 / 错别字 / 单行) | 0 文档 | 不写多余文档,直接转交执行并跑对应校验 |
| Case 2(明确局部 Bugfix,不改契约) | 0 文档 | 强制 Red-Green:先补失败复现回归测试再修绿;触及高风险须独立审查 |
| Case 3(L2 局部功能,不改外部契约) | 只建 docs/todo/ | 澄清并冻结 C1..Cn,跑 doc-plan.mjs audit 自净化直通(非平凡 Todo 建议走 Critique) |
| Case 4(L3 跨模块 / 改外部契约) | docs/specs/ + docs/todo/(按需 docs/prd/) | 跑 audit -> 1–2 轮独立只读 Critique(SELF_CRITIQUE_REJECTED 禁自审)-> 原位改净 -> critique-pass -> 交人 approve |
Case 5(与 docs/TDD.md 或生效 SPEC 冲突) | 阻断拆解 | 先修订 docs/TDD.md 并经独立确认后再续 |
说明:contributor-fork 若上游仓未接入 docs/,私有草案绝不混入 PR diff。
独立 Critique、impl 工作树协同与移交 dev-flow
- 不可自审硬锁:
doc-plan.mjs critique-pass --file ... --reviewer <id>强制校验复核人不可与文档owner相同(否则SELF_CRITIQUE_REJECTED);Critique 通过后若正文未改即可由人/维护者执行approve(修改正文会触发BLOCKED_CRITIQUE_STALE需复核)。 impl工作树协同:若flow.mjs start --task <id>已先建好隔离工作树,doc-plan.mjs init --task <id>会直接把规划单落盘到.worktrees/impl-<id>/避免撕裂;若先规划后建工作树,flow.mjs start --spec ... --todo ...会自动将规划文档迁入实施工作树并同步-issue<N>-命名。- 波次移交
dev-flow:单任务audit/approve通过后按[NEXT STEP]移交dev-flow(flow.mjs start);多切片协同可配合flow.mjs start --after <parent>起堆叠子分支或flow.mjs batch plan按文件互斥划分波次。lightweight档缺省兄弟技能时退化为原生 Git 指引。
标准命令流
# 1. 五档分流 + 门型判断 + 改哪看哪联动预警
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" triage --summary "<需求描述>" [--files <n>] [--paths a,b] [--touches-contract] [--conflicts-tdd]
# 2. 按统一模板生成文档(若已存在任务 worktree 自动落入实施工作树;自动尊重自有模板)
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" init --kind <prd|spec|todo|pair|all> --slug "<slug>" --title "<标题>" [--issue <N>] [--task "<taskId>"]
# 3. 规划内审自净化((new) 校验、Mermaid 16/12、无占位、Frontmatter 互链)
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" audit --file "docs/specs/<YYMMDD-slug-spec>.md"
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" audit .
# 4. 关联、独立 Critique、批准、成对归档(--include-master-prd 含总控 PRD)与取代
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" link "<slug-or-path>" [--issue <N>] [--pr <M>] [--branch <B>] [--status <S>]
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" critique-pass --file "docs/specs/<file>" --reviewer "<criticId>" --round 1 --summary "<结论>"
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" approve --file "docs/specs/<file>"
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" archive [<slug-or-path>] [--task <id>] [--include-master-prd]
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" supersede "<oldSpec>" --by "<newSpec>"
node ".agents/skills/dev-plan/scripts/doc-plan.mjs" listDoes 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.

