メニュー
Task generation and breakdown workflow. Use when decomposing a PRD or complex feature into implementable subtasks with clear acceptance criteria and dependencies.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/ar4mirez/samuel --skill generate-tasksコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Sync per-folder CLAUDE.md and AGENTS.md files with context-aware content. Use after structural changes, adding new directories, refactoring, or periodic maintenance to keep folder-level AI instructions current.
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Autonomous AI coding loop (Ralph Wiggum methodology). Use when running AI agents in unattended/autonomous mode to independently select and complete tasks from a structured task list (prd.json). Enables AFK development across multiple context windows.
SOC 職業分類に基づく
SKILL.md を表示中
| name | generate-tasks |
| description | Task generation and breakdown workflow. Use when decomposing a PRD or complex feature into implementable subtasks with clear acceptance criteria and dependencies. |
| license | MIT |
| metadata | {"author":"samuel","version":"1.0","category":"workflow"} |
Guide AI in creating detailed, step-by-step task list in Markdown format based on existing PRD. The task list should guide a developer through implementation with built-in quality checkpoints.
Use When: PRD created, ready to plan implementation 4D Phase: Develop (planning implementation steps)
✅ Before using this skill:
.claude/tasks/NNNN-prd-feature-name.mdCLAUDE.md exists (or will be created during this process)Phase 1: Generate parent tasks → Present to user → Wait for "Go" Phase 2: Generate detailed sub-tasks → Save file → Inform user
This ensures high-level plan approved before diving into details.
User points AI to specific PRD file.
Example:
"Generate tasks for @.claude/tasks/0001-prd-user-authentication.md using @.claude/skills/generate-tasks/SKILL.md"
AI reads and analyzes:
CRITICAL: Review existing codebase before planning.
AI must:
Check CLAUDE.md for:
Scan codebase for:
Identify reusable components:
Check guardrails compliance:
Output: Brief summary for user review before generating tasks.
Example:
## Current State Assessment
**Existing Infrastructure:**
- Express.js backend with TypeScript
- Prisma ORM with PostgreSQL
- Jest for testing (78% coverage currently)
- React frontend with Tailwind CSS
**Reusable Components:**
- `lib/validation.ts` - Input validation utilities (use Zod)
- `components/Form.tsx` - Form component (reuse for auth forms)
- `lib/errors.ts` - Centralized error handling (extend for auth errors)
**Patterns to Follow:**
- API routes in `src/api/<feature>/` structure
- Tests alongside code (`*.test.ts`)
- Repository pattern for data access
- Environment variables via `.env` with fallbacks
**Guardrails Note:**
- `src/api/user/user.controller.ts` is 285 lines (near 300 limit - don't modify)
- Create new auth module instead of extending user module
Based on PRD and current state, create high-level tasks.
Guidelines:
Save to: .claude/tasks/tasks-NNNN-prd-feature-name.md
Present to user in this format:
## High-Level Tasks
- [ ] 1.0 Database Schema & Migrations
- [ ] 2.0 Backend Authentication Service
- [ ] 3.0 API Routes & Middleware
- [ ] 4.0 Frontend Auth Components
- [ ] 5.0 OAuth Integration
- [ ] 6.0 Testing & Validation
- [ ] 7.0 Documentation & Deployment
**Relevant Files** (identified):
- `src/db/schemas/user.schema.ts` - Create
- `src/auth/auth.service.ts` - Create
- `src/api/auth/auth.controller.ts` - Create
... (full list in task file)
Then inform user:
"High-level tasks generated. Review the task structure above.
Ready to generate detailed sub-tasks? Respond with 'Go' to proceed."
PAUSE - Do not proceed until user responds with "Go" or equivalent.
This checkpoint ensures high-level plan aligns with expectations before diving into details.
Once user confirms "Go", break down each parent task into actionable sub-tasks.
Sub-task Guidelines:
Sub-task Naming Convention:
Example Parent → Sub-tasks:
- [ ] 1.0 Database Schema & Migrations
- [ ] 1.1 Create user schema with Prisma (email, password_hash, oauth_provider, created_at, updated_at)
- [ ] 1.2 Create session schema with Prisma (user_id FK, token, expires_at, created_at)
- [ ] 1.3 Generate and run Prisma migration for user and session tables
- [ ] 1.4 Verify schema in PostgreSQL (check indexes, constraints)
- [ ] 1.5 Update .env.example with DATABASE_URL placeholder
For each sub-task, identify which guardrails must be validated.
Format: Add guardrail checklist as comment under relevant tasks.
Example:
- [ ] 2.3 Implement password hashing service (bcrypt, cost factor 12)
<!-- Guardrails:
✓ All environment variables have secure defaults (BCRYPT_ROUNDS=12)
✓ All exported functions have type signatures and JSDoc
✓ Function ≤50 lines (hash, verify, validate strength as separate functions)
✓ Edge cases tested (null, empty, too short, invalid chars)
-->
For each sub-task, estimate complexity in tokens (not time).
Complexity Levels:
Example:
- [ ] 2.3 Implement password hashing service [~3,000 tokens - Medium]
List all files that will be created or modified, including tests.
Format: See references/process.md for detailed file listing format.
Provide guidance for each task section.
Include:
Example:
## Implementation Notes
### Testing Requirements
- All auth logic must have >95% coverage (business-critical)
- Test files alongside code (e.g., `auth.service.ts` + `auth.service.test.ts`)
- Use `npm test src/auth` to run auth module tests only
- Integration tests must test actual JWT validation, not mocks
### Security Checklist (verify before each commit)
- [ ] All passwords hashed (never store plain text)
- [ ] All JWT secrets from environment variables
- [ ] All inputs validated with Zod schemas
- [ ] All SQL queries parameterized (use Prisma)
- [ ] Rate limiting on login/register endpoints
- [ ] Authentication events logged
### Performance Targets
- Login/register: <200ms (p95)
- JWT validation: <50ms (p95)
- Password hashing: 100-300ms (bcrypt cost 12)
### Common Pitfalls
- ❌ Don't validate JWT in database query (use middleware)
- ❌ Don't store tokens in localStorage (use httpOnly cookies)
- ❌ Don't return sensitive data in error messages
- ❌ Don't skip rate limiting (prevents brute force)
Combine all sections into final task list structure.
See references/process.md for complete output format template.
File: .claude/tasks/tasks-NNNN-prd-feature-name.md
Format: Markdown
Naming: Match PRD filename (e.g., PRD is 0001-prd-user-auth.md → tasks are tasks-0001-prd-user-auth.md)
User will say: "Start on task 1.1" or "Continue with next task"
AI should:
feat(auth): task 1.1 - create user schemaFor each sub-task, identify which of CLAUDE.md's 30+ guardrails apply.
Common guardrails per task type:
Phase 1 (High-Level):
Phase 2 (Detailed):
During Implementation:
Primary: Junior developer with AI assistance Assumption: Developer knows fundamentals but needs clear guidance on:
See references/process.md for:
Remember: Good task breakdown = clear roadmap. Each sub-task should be:
If the project uses autonomous AI coding loops (Ralph Wiggum methodology), the generated task list can be converted to machine-readable format for unattended execution.
After generating the markdown task list:
samuel auto init --prd .claude/tasks/NNNN-prd-feature.md
This converts the PRD and task list into .claude/auto/prd.json and
generates the loop orchestration files. See .claude/skills/auto/SKILL.md
for the full autonomous workflow.