.claude/commands/reflection.md
code:reflection.md
---
description: Kolb's experiential learning cycle for extracting reusable principles from session experience
---
Transform session experience into reusable principles through Kolb's Learning Cycle.
**Framework: Kolb's Experiential Learning Model**
Concrete Experience → Reflective Observation → Abstract Conceptualization → Active Experimentation
**Critical Constraint**: Phase 3 (Abstract Conceptualization) is MANDATORY.
No concrete implementation details may enter CLAUDE.md without abstraction.
**Output Language**: ALWAYS use the same language as the user's session (e.g., Japanese session → Japanese output).
---
## Phase 1: Concrete Experience
**Purpose**: Organize session logs as factual records. NO analysis or interpretation.
**Critical**: Focus ONLY on facts. NO analysis, interpretation, or reflection in this phase.
**Structure**: Record facts for each of the 3 axes.
**Output Format**:
`
## Phase 1: Concrete Experience
### Goal Achievement Facts
### Efficiency Facts
### User Satisfaction Facts
`
---
## Phase 2: Reflective Observation
**Purpose**: Analyze patterns and root causes from experience.
**Tasks**:
- 3-axis evaluation (Goal Achievement / Efficiency / User Satisfaction)
- Pattern recognition: What repeated?
- Root cause analysis: Why did it happen? (3 Whys)
**Output Format**:
`
## Phase 2: Reflective Observation
### Goal Achievement: 7/10
### Efficiency: 6/10
### User Satisfaction: 8/10
### Pattern Recognition
What repeated during this session:
`
---
## Phase 3: Abstract Conceptualization
**Purpose**: Extract reusable principles from concrete experience.
🚨 **CRITICAL: Abstraction Quality Gate** 🚨
Every principle MUST pass this checklist:
`
□ Removed concrete implementation details (function names, file names, URLs, etc.)
□ Specified "Apply when" conditions (when to apply)
□ Described as reusable principle (applicable to other projects)
□ Checked for duplication/conflict with existing CLAUDE.md rules
□ Verified applicability even in counter-examples
`
**Tasks**:
1. Create Concrete→Abstract Transformation Table
2. Extract reusable principles
3. Verify abstraction quality
4. Check integration with existing CLAUDE.md rules
**Output Format**:
`
## Phase 3: Abstract Conceptualization
### Extracted Principles
`
---
## Phase 4: Active Experimentation
**Purpose**: Propose practical improvements to increase success reproducibility and prevent failure recurrence.
**Improvement Types**:
- Memory file (CLAUDE.md) updates: Add/modify/delete context and rules
- External scripts: Automate deterministic solutions
- Lint rules: Enforce patterns automatically
- Other tools: Any external mechanisms to prevent recurrence
**Critical**: Present all improvements as numbered proposals. User approves by specifying proposal numbers.
**Output Format**:
`
## Phase 4: Active Experimentation
### Improvement Proposals
- Implementation:
`markdown
`
- Implementation:
`bash
`
- Implementation:
`json
`
---
**Approve proposals** (specify numbers, e.g., "1,3" or "all" or "none"):
`
---
## Guidelines
- **Thorough Abstraction**: Never add concrete implementation details to CLAUDE.md
- **Reusability**: Formulate in a way applicable to other projects and contexts
- **Consistency with Existing Rules**: Align with existing CLAUDE.md format
- **Verifiability**: Make principles verifiable in next session
---
## Anti-patterns
❌ **Bad**: Verify email validation before calling UserService.createUser()
✅ **Good**:
`markdown
### Input Validation Protocol
**Apply when**: Processing user input in any function/endpoint
- Verify validation logic executes beforehand
- Verify error handling on failure
`
❌ **Bad**: POST /api/users must return 201
✅ **Good**:
`markdown
### HTTP Status Code Consistency
**Apply when**: RESTful API design/implementation
- Use 201 for successful resource creation
- Verify idempotency and status code correspondence
`