to-plan

Files (1)

• SKILL.md

SKILL.md (first lines)

# To Plan

把 PRD、Spec、release goal 或 conversation context 转成执行计划。Plan 只决定阶段、节奏、依赖、风险和文档同步点，不写功能契约，也不拆最小任务。

## Workflow

1. **Load repo workflow**
   - Read the repo's actual agent adapter if present, such as `AGENTS.md`, `CLAUDE.md`, `.cursor/rules/`, or `.github/copilot-instructions.md`
   - Read `docs/agents/document-system.md` and `docs/agents/domain.md` if they exist
   - Read `CONTEXT.md` and relevant `docs/adr/`
   - If BMad is involved, read `docs/BMAD_INTEGRATION.md` and `docs/LEVEL3_BMAD_WORKFLOW.md`

2. **Find source material**
   - Prefer `docs/features/<feature>/PRD.md`
   - Also read `docs/features/<feature>/SPEC.md` if it exists
   - For release or cross-feature work, read relevant files under `docs/features/`
   - If BMad was used, read relevant epics/stories, sprint planning, readiness, or architecture output as sequencing source material
   - If source material is only conversation context, mark assumptions explicitly

3. **Choose plan location**
   - Cross-feature, cross-stage, or release plan: `docs/plans/<plan-slug>.md`
   - Single large feature plan: `docs/features/<feature>/PLAN.md`
   - If one `SPEC.md` plus 3-5 issues is enough, do not create a Plan; say it should be skipped

4. **Draft the plan**
   - Define phases, goals, deliverables, acceptance standards, dependencies, risks, and document sync points
   - Keep phases outcome-oriented, not layer-by-layer implementation
   - Reference PRD/Spec/ADR paths explicitly
   - If BMad story cycle will own execution, do not duplicate every story as issue-level plan work; record only phase sequencing, sync points, readiness gates, and when to enter `bmad-sprint-planning`
   - If AI-native issues will own execution, make the plan clear enough for `/to-issues` to create vertical slices

5. **Write or update**
   - Create the chosen Plan file
   - If updating an existing Plan, preserve completed phases and append changes instead of rewriting history

6. **Finish with next step**
   - If behavior or API is not specified, suggest `/to-spec`
   - If the Spec is ready and AI-native issues own execution, suggest `/to-issues`
   - If BMad story cycle owns execution, suggest final docs lock / readiness resolution, then `bmad-sprint-planning`; use `/to-issues` only for control-panel issues
   - If terminology or durable decisions are unresolved, suggest `/grill-with-docs`

## Required Plan Shape

```markdown
# Plan: <Plan Name>

## Overview

## Source Documents

- PRD: docs/features/<feature>/PRD.md
- Spec: docs/features/<feature>/SPEC.md or None yet
- ADR: docs/adr/<adr-file>.md or None

## Phases

### Phase 1: <Name>

- **Goal**:
- **Deliverables**:
- **Acceptance Standard**:
- **Dependencies**:
- **Risks**:

### Phase 2: <Name>

## Document Sync Points

## Open Questions
```

## Quality Bar

- Plan is optional; create it only when sequencing is genuinely useful.
- Plan does not define API contracts, data models, state machines, or validation rules.
- Plan does not create issue-level task lists; it names deliverables and sequencing.
- Each phase has a clear completion signal.
- Risks include mitigation or the next decision needed.

From pop1414/ai-native-doc-system, MIT.

More from pop1414/ai-native-doc-system