| name | reflect |
| description | Reflect on a completed coding iteration to identify improvements |
Iteration Reflection Skill
Use this skill at the END of each coding iteration to reflect on what happened and identify improvements.
Reflection Framework
After completing an iteration, systematically evaluate:
1. What Went Well? ✅
Identify positive patterns to reinforce:
- Did the task selection work well?
- Was the component well-defined?
- Did tests provide good coverage?
- Was the code quality high?
- Did documentation help?
- Were dependencies minimal?
Example from JSONL Logger iteration:
- ✅ AGENTS.md pattern worked perfectly - quick orientation
- ✅ Component selection excellent (independent, foundational, testable)
- ✅ Zero clippy warnings, comprehensive tests
- ✅ Clean module organization
2. What Needed Guidance? ⚠️
Identify friction points that required user intervention:
- Were there ambiguities in requirements?
- Did I make incorrect assumptions?
- Were there preference mismatches?
- Did I need correction on approach?
- Were there multiple back-and-forth interactions for simple things?
Example from JSONL Logger iteration:
- ⚠️ Git attribution preference - required correction + settings update (2 interactions)
- ⚠️ Not pure TDD - wrote tests + implementation together instead of RED→GREEN→REFACTOR
3. Feedback Automation Analysis 🔧
Evaluate the development feedback loop:
Questions to ask:
- Were there manual steps that could be automated?
- Did I forget to run important checks?
- Could errors have been caught earlier?
- Are there repetitive commands that should be scripted?
- Is the feedback loop fast enough?
- Are there missing guardrails?
Example issues identified:
- ❌ No pre-commit verification - easy to commit broken code
- ❌ Manual process - have to remember all verification commands
- ❌ No continuous feedback for TDD
- ❌ Didn't verify tooling before using it
4. Concrete Action Items 📋
Critical: Don't just identify problems - propose concrete, actionable solutions.
Format each action item as:
Priority: [High/Medium/Low]
Problem: [What's broken or missing]
Solution: [Specific, implementable fix]
Effort: [Small/Medium/Large]
Example action items from JSONL Logger iteration:
Priority: High
Problem: No pre-commit verification - can commit broken code
Solution: Add .git/hooks/pre-commit that runs tests, clippy, fmt
Effort: Small
Priority: High
Problem: Manual verification is error-prone
Solution: Document automated pre-commit hook
Effort: Small
Priority: Medium
Problem: TDD workflow not clearly defined
Solution: Create /coding-iteration skill with RED→GREEN→REFACTOR
Effort: Medium
Priority: Low
Problem: Suggested cargo-watch but doesn't fit my execution model
Solution: Skip watch mode (doesn't make sense for blocking tools)
Effort: None
Reflection Template
Copy this template for each iteration reflection:
## Iteration: [Component Name]
### ✅ What Went Well
-
-
-
### ⚠️ What Needed Guidance
-
-
### 🔧 Feedback Automation Analysis
Current gaps:
-
Potential improvements:
-
### 📋 Action Items
**High Priority:**
1. [Problem] → [Solution] (Effort: X)
2.
**Medium Priority:**
1. [Problem] → [Solution] (Effort: X)
**Low Priority / Future:**
1.
### Action Item Disposition (ALL items must be addressed)
| # | Item | Disposition | Artifact |
|---|------|-------------|----------|
| 1 | ... | Implemented / Added to TODO.md / Added to AGENTS.md | link or file |
| 2 | ... | ... | ... |
When to Reflect
Trigger reflection when:
- Completing a significant component
- User provides corrective feedback
- Encountering repeated friction
- Finishing a multi-step task
- User explicitly asks "/reflect"
Don't reflect on:
- Trivial changes (typo fixes, formatting)
- Middle of a flow (wait until component done)
- When explicitly told to skip it
Acting on Reflections
Not every reflection produces action items — sometimes things went well and there's nothing to change. That's fine.
But if action items arise, they must be dealt with immediately. Don't list them and move on.
For each action item, do one of:
- Implement now: Make the change (update docs, add validation, fix the process)
- Track in TODO.md: Add as a task with priority, description, and context
- Add to AGENTS.md: If it's a process/anti-pattern lesson
No action item should be left unaddressed. Before ending the reflection, verify every item was either implemented or tracked in a file.
Meta-Improvement
This skill itself should improve over time:
- If reflection finds consistent patterns, update this skill
- If certain categories are never useful, remove them
- If new types of issues emerge, add categories
- Keep it practical - reflection should take < 5 min
Example: Complete Reflection
## Iteration: JSONL Logger Implementation
### ✅ What Went Well
- AGENTS.md structure provided perfect orientation
- Task selection was optimal (independent, foundational)
- 6 comprehensive tests with 100% coverage
- Zero clippy warnings achieved
- Clean separation of concerns
### ⚠️ What Needed Guidance
- Git attribution: didn't know user preference, needed 2 interactions to fix
- TDD process: wrote tests + impl together instead of RED first
### 🔧 Feedback Automation Analysis
- Missing: Pre-commit validation hook
- Missing: Automated verification before commit
- Missing: Clear TDD workflow documentation
- Issue: Suggested tools (cargo-watch) that don't fit my execution model
### 📋 Action Items
**High Priority:**
1. Add pre-commit hook → Auto-run tests/clippy/fmt (Small)
2. Create /coding-iteration skill → Document proper TDD (Medium)
3. Update settings.json → Disable git attribution (Small)
**Medium Priority:**
1. Update AGENTS.md → Reference new automation (Small)
**Low Priority:**
1. cargo-watch → Skip (doesn't fit blocking execution model)
### Decision: Implement now before next iteration
Remember: A clean reflection with no action items is a sign things are going well. But if problems surface, deal with them — don't just list them.