writing-plans

Create detailed implementation plans before coding (saves tokens later)

$ Installer

git clone https://github.com/LEEI1337/phantom-neural-cortex /tmp/phantom-neural-cortex && cp -r /tmp/phantom-neural-cortex/.claude/skills/writing-plans ~/.claude/skills/phantom-neural-cortex

// tip: Run this command in your terminal to install the skill


name: writing-plans description: Create detailed implementation plans before coding (saves tokens later) trigger: auto

Writing Implementation Plans

Create clear plans BEFORE coding to save time and tokens:

Plan Structure

1. GOAL (1-2 sentences)

What are we building and why?

2. CONSTRAINTS

- Technology limits
- Time/resource constraints
- Compatibility requirements
- Performance targets

3. APPROACH (High-Level)

- Architecture pattern
- Key components
- Data flow
- Integration points

4. STEPS (Ordered Tasks)

1. [Component] - Brief description
2. [Component] - Brief description
   - Sub-task if needed
3. [Testing] - What to test
4. [Deployment] - How to deploy

5. RISKS & MITIGATIONS

Risk: Description
Mitigation: Solution

6. SUCCESS CRITERIA

- Measurable outcomes
- Test coverage targets
- Performance metrics

Token Optimization

  • Write plan ONCE, reference during implementation
  • Clear steps prevent back-and-forth
  • Catches issues early (cheaper to fix in planning)
  • Team can review before coding starts

Example Output Format

GOAL: Add user authentication with JWT

CONSTRAINTS:
- Must use existing DB schema
- Response time < 200ms
- Support OAuth2

APPROACH:
- Middleware-based auth layer
- JWT with refresh tokens
- Redis session cache

STEPS:
1. Create auth middleware
2. Implement token generation/validation
3. Add refresh token logic
4. Write integration tests
5. Update API docs

RISKS:
- Token expiry handling → Use refresh tokens
- Concurrency issues → Use Redis locks

SUCCESS:
- All endpoints secured
- 95%+ test coverage
- Load test passes (1000 req/s)