一键导入
blueprint
Mandatory planning gate. Decomposes any task into phases, dependencies, risks, and success criteria before any implementation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Mandatory planning gate. Decomposes any task into phases, dependencies, risks, and success criteria before any implementation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Bun runtime: HTTP server, file I/O, SQLite, test runner, package manager, bundler — all-in-one JS toolchain.
Clerk: Drop-in auth UI, Organizations, User management, JWT templates, webhooks, Next.js middleware integration.
Gelişmiş masaüstü, tarayıcı ve işletim sistemi kontrol yeteneği. Görsel (koordinat tabanlı) fare/klavye otomasyonu, DOM manipülasyonu, pencere yönetimi, gelişmiş dosya, ağ ve süreç yönetimini kapsar.
Drizzle ORM: Schema definition, type-safe queries, migrations, relations, Postgres/SQLite/MySQL support.
Expo Router v3: File-based navigation, layouts, tabs, modals, deep linking, API routes, typed routes.
Flutter ile oyun geliştirme (Flame Engine vb.) ve karmaşık, büyük ölçekli mimariler kurma rehberi.
| name | blueprint |
| description | Mandatory planning gate. Decomposes any task into phases, dependencies, risks, and success criteria before any implementation. |
| triggers | {"keywords":["plan","blueprint","approach","architecture","design","decompose","how should","strategy"]} |
| auto_load_when | Before any implementation task with 2+ steps (mandatory gate) |
| agent | planner |
| tools | ["Read","Write","Bash"] |
Rule: No implementation without a blueprint. This is the non-negotiable first step.
How complex is this?
├── 1 file, 1 concern, <30 min → SIMPLE — no blueprint needed, proceed directly
├── 2-4 files, clear scope → MEDIUM — quick blueprint (5 min)
└── 5+ files, multi-agent → COMPLEX — full DAG blueprint required
If MEDIUM or COMPLEX → continue below
Fill this for every MEDIUM/COMPLEX task:
## Blueprint: {task name}
### Goal
One sentence: what success looks like.
### Scope
- Files affected: [list]
- New dependencies: [list or none]
- Breaking changes: yes/no
### Phases
#### Phase 1: Research (parallel)
- [ ] Read existing code patterns
- [ ] Check for similar implementations
- [ ] Identify constraints
#### Phase 2: Implementation
- [ ] Step A (depends on: nothing)
- [ ] Step B (depends on: A)
- [ ] Step C (depends on: A)
#### Phase 3: Verification
- [ ] Tests pass
- [ ] TypeScript check clean
- [ ] Security: no new vulnerabilities
- [ ] Lint: no new warnings
### Risks
| Risk | Likelihood | Mitigation |
|---|---|---|
| Breaking existing API | Medium | Add to existing, don't replace |
| Performance regression | Low | Benchmark before/after |
### Definition of Done
- [ ] All Phase 3 checks pass
- [ ] Deployed and verified in staging
Proceed? (Yes / Adjust)
For multi-agent tasks, express dependencies explicitly:
task-001: researcher — Investigate existing patterns [depends: nothing]
task-002: architect — Design component structure [depends: 001]
task-003: frontend-ops — Implement UI [depends: 002]
task-004: style-arch — Apply design tokens [depends: 003]
task-005: qa-specialist — Write tests [depends: 002] ← parallel with 003
task-006: security — Review auth changes [depends: 003]
task-007: docs-agent — Update documentation [depends: 003, 004]
Phases:
Risk score = Likelihood × Impact
Likelihood: Low=1, Medium=2, High=3
Impact: Low=1, Medium=2, High=3
Score ≥ 6 → add explicit mitigation step to Phase 1
Score ≥ 8 → present to user before proceeding
Phase 1: Read existing patterns, identify extension points
Phase 2: Implement (server) → implement (client) → connect
Phase 3: Unit test → integration test → lint → typecheck
Phase 1: Reproduce → identify root cause → find all affected locations
Phase 2: Fix root cause → update tests
Phase 3: Verify fix → check for regressions
Phase 1: Map all usages → identify safe extraction boundaries
Phase 2: Extract → update imports → verify types
Phase 3: All existing tests pass → no behavior change
Phase 1: Design request/response shape → check auth requirements
Phase 2: Route handler → validation → business logic → tests
Phase 3: Security review → rate limiting → documentation
❌ Starting implementation immediately without stating the plan
✅ Always state the plan first, even for "simple" tasks
❌ Blueprinting in your head without writing it down
✅ Write the blueprint — it forces clarity and catches gaps
❌ Skipping Phase 3 (verification) because "it looks right"
✅ Verification is non-negotiable — it catches silent failures
❌ 10-step blueprints that cover everything
✅ Max 7 tasks per blueprint — split if larger
Always present blueprint as:
[BLUEPRINT] Task: {name}
Complexity: Simple / Medium / Complex
Phases: N | Agents: [list] | Estimated tasks: N
[Phase summary]
Proceed? (Yes / Adjust / Simplify)