微信扫一扫Reflect - Agent Self-Improvement Skill
Transform your AI assistant into a continuously improving partner. Every correction becomes a permanent improvement that persists across all future sessions.
Quick Reference
| Command | Action |
|---------|--------|
| reflect | Analyze conversation for learnings |
| reflect on | Enable auto-reflection |
| reflect off | Disable auto-reflection |
| reflect status | Show state and metrics |
| reflect review | Review pending learnings |
When to Use
- After completing complex tasks
- When user explicitly corrects behavior ("never do X", "always Y")
- At session boundaries or before context compaction
- When successful patterns are worth preserving
Workflow
Step 1: Scan Conversation for Signals
Analyze the conversation for correction signals and learning opportunities.
Signal Confidence Levels:
| Confidence | Triggers | Examples | |------------|----------|----------| | HIGH | Explicit corrections | "never", "always", "wrong", "stop", "the rule is" | | MEDIUM | Approved approaches | "perfect", "exactly", "that's right", accepted output | | LOW | Observations | Patterns that worked but not explicitly validated |
See signal_patterns.md for full detection rules.
Step 2: Classify & Match to Target Files
Map each signal to the appropriate target:
| Category | Target Files |
|----------|--------------|
| Code Style | code-reviewer, backend-developer, frontend-developer |
| Architecture | solution-architect, api-architect, architecture-reviewer |
| Process | CLAUDE.md, orchestrator agents |
| Domain | Domain-specific agents, CLAUDE.md |
| Tools | CLAUDE.md, relevant specialists |
| New Skill | Create new skill file |
See agent_mappings.md for mapping rules.
Step 3: Check for Skill-Worthy Signals
Some learnings should become new skills rather than agent updates:
Skill-Worthy Criteria:
- Non-obvious debugging (>10 min investigation)
- Misleading error (root cause different from message)
- Workaround discovered through experimentation
- Configuration insight (differs from documented)
- Reusable pattern (helps in similar situations)
Quality Gates (must pass all):
- [ ] Reusable: Will help with future tasks
- [ ] Non-trivial: Requires discovery, not just docs
- [ ] Specific: Can describe exact trigger conditions
- [ ] Verified: Solution actually worked
- [ ] No duplication: Doesn't exist already
Step 4: Generate Proposals
Present findings in structured format:
# Reflection Analysis
## Session Context
- **Date**: [timestamp]
- **Messages Analyzed**: [count]
## Signals Detected
| # | Signal | Confidence | Source Quote | Category |
|---|--------|------------|--------------|----------|
| 1 | [learning] | HIGH | "[exact words]" | Code Style |
## Proposed Changes
### Change 1: Update [agent-name]
**Target**: `[file path]`
**Section**: [section name]
**Confidence**: HIGH
```diff
+ New rule from learning
Review Prompt
Apply these changes? (Y/N/modify/1,2,3)
### Step 5: Apply with User Approval
**On `Y` (approve):**
1. Apply each change using Edit tool
2. Commit with descriptive message
3. Update metrics
**On `N` (reject):**
1. Discard proposed changes
2. Log rejection for analysis
**On `modify`:**
1. Present each change individually
2. Allow editing before applying
**On selective (e.g., `1,3`):**
1. Apply only specified changes
2. Commit partial updates
## State Management
State is stored in `~/.reflect/` (configurable via `REFLECT_STATE_DIR`):
```yaml
# reflect-state.yaml
auto_reflect: false
last_reflection: "2026-01-26T10:30:00Z"
pending_reviews: []
Metrics Tracking
# reflect-metrics.yaml
total_sessions_analyzed: 42
total_signals_detected: 156
total_changes_accepted: 89
acceptance_rate: 78%
confidence_breakdown:
high: 45
medium: 32
low: 12
most_updated_agents:
code-reviewer: 23
backend-developer: 18
skills_created: 5
Safety Guardrails
Human-in-the-Loop
- NEVER apply changes without explicit user approval
- Always show full diff before applying
- Allow selective application
Incremental Updates
- ONLY add to existing sections
- NEVER delete or rewrite existing rules
- Preserve original structure
Conflict Detection
- Check if proposed rule contradicts existing
- Warn user if conflict detected
- Suggest resolution strategy
Output Locations
Project-level (versioned with repo):
.claude/reflections/YYYY-MM-DD_HH-MM-SS.md- Full reflection.claude/skills/{name}/SKILL.md- New skills
Global (user-level):
~/.reflect/learnings.yaml- Learning log~/.reflect/reflect-metrics.yaml- Aggregate metrics
Examples
Example 1: Code Style Correction
User says: "Never use var in TypeScript, always use const or let"
Signal detected:
- Confidence: HIGH (explicit "never" + "always")
- Category: Code Style
- Target:
frontend-developer.md
Proposed change:
## Style Guidelines
+ * Use `const` or `let` instead of `var` in TypeScript
Example 2: Process Preference
User says: "Always run tests before committing"
Signal detected:
- Confidence: HIGH (explicit "always")
- Category: Process
- Target:
CLAUDE.md
Proposed change:
## Commit Hygiene
+ * Run test suite before creating commits
Example 3: New Skill from Debugging
Context: Spent 30 minutes debugging a React hydration mismatch
Signal detected:
- Confidence: HIGH (non-trivial debugging)
- Category: New Skill
- Quality gates: All passed
Proposed skill: react-hydration-fix/SKILL.md
Troubleshooting
No signals detected:
- Session may not have had corrections
- Check if using natural language corrections
Conflict warning:
- Review the existing rule cited
- Decide if new rule should override
- Can modify before applying
Agent file not found:
- Check agent name spelling
- May need to create agent file first