dev-planner
暂无简介
- License 许可
- MIT
- 所属仓库
- zxpmail/ReqForge
- 最近更新
- 2026-10-07
- Stars
- ★ 18
- Forks
- 5
02 资源文件 11
- 参考文档commands/dev-planner.md503 B
- 参考文档references/analysis-dimension-checklist.md3.2 KB
- 参考文档references/analysis-strategies.md4.5 KB
- 参考文档references/architecture-health-pass.md1.3 KB
- 参考文档references/first-principles.md3.7 KB
- 参考文档references/generate-contracts.md3.6 KB
- 参考文档references/plan-critique-check.md5.0 KB
- 参考文档references/plan-hard-gate-rationalization.md866 B
- 参考文档references/workflow.md13.4 KB
- 参考文档skill.json936 B
- 参考文档templates/dev-plan-template.md9.3 KB
03 SKILL.md 正文
<!-- forge: dev-planner v1.1 -->
name: dev-planner description: Used when Product-Spec.md is complete and needs to be planned into development phases. Also used to update existing development plans after Spec changes. Outputs DEV-PLAN.md. version: 1.1.0 updated: 2026-05-26 requires: []
<!-- begin: task --> [Task] Generation Mode: Read Product-Spec.md (and Design-Brief.md, if present), analyze feature dependency relationships, WebSearch to validate technology choices, output a phased development plan DEV-PLAN.md.
**Iteration Mode**: When the Product Spec changes, analyze the scope of impact, update the Phase breakdown and file inventory in DEV-PLAN.md. Completed Phases (marked with [x]) remain untouched.
<!-- end: task --> <!-- begin: not-for --> [Not For] - Writing actual code -> use /dev-builder instead - Gathering requirements -> use /product-spec-builder instead - Fixing bugs -> use /bug-fixer instead
<!-- end: not-for --> <!-- begin: dependency-check --> [Dependency Check] Executed automatically as the first step when the Skill starts:
Required:
- Product-Spec.md -> if missing, prompt user to call /product-spec-builder first
- Product-Spec.md must include completed **§ Idea Stage Exit Criteria** (three questions) — if missing or `TBD`, route back to `/product-spec-builder` before generating DEV-PLAN
Optional (degradation mode):
- Design-Brief.md -> if missing, mark as "no design specification mode", visual details annotated as [TBD by Design Brief]
- Design tool MCP -> if not connected or no files, rely only on text descriptions, mark as "no design draft mode"
- Existing project code -> if present, scan existing structure as constraints, enter iteration mode
<!-- end: dependency-check -->
<!-- begin: first-principles -->
[First Principles]
Plan 前必读 references/first-principles.md
Tech Stack: 规划前读 .forge/dev-map.md 的技术栈节,确保 DEV-PLAN.md 的技术栈与 dev-map 一致。
<!-- end: first-principles -->
<!-- begin: shared-discipline -->
[Shared Discipline]
Karpathy 四原则 → ../_shared/karpathy-discipline.md
<!-- end: shared-discipline -->
<!-- begin: hard-gate -->
[HARD-GATE]
Until DEV-PLAN.md saved and user explicitly confirms → write .forge/plan-confirmed.json → MUST NOT invoke /dev-builder. Chat "looks good" ≠ confirm.
Prerequisites: Product-Spec.md must exist. Rationalizations → references/plan-hard-gate-rationalization.md
<!-- end: hard-gate -->
<!-- begin: file-structure -->
[File Structure]
dev-planner/ ├── SKILL.md ├── commands/dev-planner.md └── references/ ├── first-principles.md ├── analysis-dimension-checklist.md ├── analysis-strategies.md ├── workflow.md ├── architecture-health-pass.md └── plan-hard-gate-rationalization.md ../_shared/
<!-- end: file-structure -->
<!-- begin: output-style -->
[Output Style]
→ ../_shared/output-style-concise.md(Plan 额外要求:每 Phase 可独立编译运行;Task 列具体文件路径;禁止 TBD)
→ DEV-PLAN 完成必须附加 ../_shared/output-status-protocol.md
<!-- end: output-style -->
<!-- begin: gotchas -->
[Gotchas]
Unrealistic Phasing: Each Phase must produce compilable, runnable output — split if no visible outcome.
Missing dependency order: infrastructure → data → API → UI.
Tech stack without WebSearch: Confirm versions before writing DEV-PLAN.
Ignoring existing code: Iteration mode — scan structure first.
Missing MVP Scope: Fill ## MVP Scope before Phase 1.
<!-- end: gotchas -->
<!-- begin: output-artifacts -->
[Output Artifacts]
- DEV-PLAN.md — Phased development plan (created in generation mode, updated in iteration mode)
- changes/<change-name>/tasks.md — Task breakdown (filled when /change-manager apply invokes dev-planner for that change only — not by product-spec-builder iteration)
- .forge/content-verify.json — (optional, recommended) forge-verify contract base generated from Phase requirements. Produced by Step 6 of Output Phase when user confirms plan. See references/generate-contracts.md.
<!-- end: output-artifacts --> <!-- begin: analysis-dimension-checklist --> [Dimension Checklist] See references/analysis-dimension-checklist.md for the full analysis dimension checklist.
Must-have dimensions:
- **Technology Stack**: framework + version + key deps confirmed
- **Phase Breakdown**: ordered sequence based on dependency relationships
- **Delivery Checklist**: each Phase has verifiable deliverables
- **Key Files**: each Phase lists specific file paths
- **Dependency Graph**: Phase ordering respects feature dependencies
[Anti-Rationalization Checklist]
→ references/plan-hard-gate-rationalization.md
在 DEV-PLAN.md 确认前禁止调用 /dev-builder。
[Plan Critique Check]
→ references/plan-critique-check.md
Generation Mode: Analysis Phase 后、Output Phase 前执行。挑战 Phase 顺序、MVP 范围、技术栈选择。
Iteration Mode / 用户说 "skip plan critique" → 跳过。
[Analysis Dimension Checklist]
Planning 前读取 references/analysis-dimension-checklist.md
<!-- end: analysis-dimension-checklist -->
<!-- begin: analysis-strategies -->
[Analysis Strategies]
按需读取 references/analysis-strategies.md
<!-- end: analysis-strategies -->
<!-- begin: information-sufficiency-criteria -->
[Information Sufficiency Criteria]
Must satisfy before output: tech stack verified, Phase breakdown + key files + dependency order, all Spec features covered. Details in references/analysis-dimension-checklist.md §Information Sufficiency.
<!-- end: information-sufficiency-criteria --> <!-- begin: quality-rubric --> [Quality Rubric] 10-item, 20 point scoring system. Ship threshold: >= 16 with no critical item scoring 0.
| # | Dimension | Pts | Critical | Scoring |
|---|-----------|-----|----------|---------|
| 1 | Phase independence | 2 | YES | 2 = Each Phase produces compilable, runnable, testable output; 1 = Most Phases deliverable but one depends on future phases; 0 = Phases are just task groupings with no runnable milestone |
| 2 | Dependency ordering | 2 | YES | 2 = Phases follow correct technical order: infra -> data -> API -> UI; 1 = Minor ordering issues but non-blocking; 0 = Backend phase scheduled after frontend that depends on it |
| 3 | MVP clarity | 2 | YES | 2 = Phase 1 is a true MVP with working core functionality, not just setup; 1 = Phase 1 is mostly boilerplate with minimal functionality; 0 = Phase 1 is pure project init with no user-visible output |
| 4 | Task granularity | 2 | -- | 2 = Tasks small enough for one session (2-4 hours); concrete deliverables; 1 = Some tasks reasonable, others too large; 0 = Tasks are vague epics with no clear completion point |
| 5 | Key file completeness | 2 | -- | 2 = Every Phase lists concrete files to create/modify with paths; 1 = Most Phases have file lists but some missing; 0 = No file-level detail, just "implement X" |
| 6 | Acceptance criteria | 2 | YES | 2 = Each Phase has verifiable completion criteria (testable); 1 = Criteria present but some vague or untestable; 0 = No acceptance criteria beyond "works correctly" |
| 7 | Risk awareness | 2 | -- | 2 = Identifies risky items (third-party deps, complex algorithms, perf) with mitigation; 1 = Some risks mentioned but no mitigation; 0 = No risk assessment |
| 8 | Technical stack alignment | 2 | -- | 2 = Plan respects chosen stack constraints; versions verified via WebSearch; 1 = Mostly aligned but some tasks assume unavailable APIs; 0 = Plan contradicts chosen tech stack |
| 9 | Estimability | 2 | -- | 2 = Each Phase has rough time estimate; cross-phase dependencies explicit; 1 = Estimates present but not tied to task complexity; 0 = No estimates or dependency mapping |
| 10 | Iteration responsiveness | 2 | -- | 2 = Plan cleanly accommodates Spec changes without invalidating completed Phases; 1 = Minor disruption from changes but manageable; 0 = Single Spec change forces full replan |
**Scoring**: Run `pnpm validate-skill --score core/skills/dev-planner` to compute.
<!-- end: quality-rubric -->
<!-- begin: workflow -->
[Workflow]
1. Run [Dependency Check]
2. Read references/first-principles.md
3. 必须先 Read references/workflow.md 对应模式章节(Generation / Iteration),再输出 DEV-PLAN
4. Apply references/analysis-dimension-checklist.md + references/analysis-strategies.md during Analysis
5. Optional: references/architecture-health-pass.md when user requests architecture health review
<!-- end: workflow -->
<!-- begin: architecture-health-pass -->
[Architecture Health Pass]
Optional → references/architecture-health-pass.md
<!-- end: architecture-health-pass --> <!-- begin: initialization --> [Initialization] Generation Mode if no DEV-PLAN.md (or greenfield); Iteration Mode if DEV-PLAN exists + Spec changed. Route via [Workflow] step 3.
<!-- end: initialization -->
04 同仓库更多 Skills
feedback-writer
Analyze and record user corrections, feedback, and behavior preferences as structured feedback files for the evolution engine.
evolution-engine
Scan accumulated feedback to identify patterns, generate evolution proposals for SKILL.md, rules, or new Skills.
db-migration-checker
暂无简介
skill-builder
暂无简介
request-dispatcher
暂无简介
reqforge-greenkeeper
暂无简介