haddock-development/claude-reflect-system / reflect
reflect
UnverifiedAnalyzes conversation transcripts to extract user corrections, patterns, and preferences, then proposes skill improvements. Use this skill when users provide corrections, express preferences about code style, or when patterns emerge from successful approaches. Can be triggered manually with /reflect or automatically at session end when enabled.
Reflect - Self-Improving Skills
Overview
This skill enables Claude Code to learn from conversations by analyzing corrections, approvals, and patterns, then proposing updates to relevant skills. It implements a "correct once, never again" learning system.
Usage Modes
1. Manual Reflection (/reflect)
Trigger analysis of the current conversation:
/reflect [skill-name]
- Without skill-name: Analyzes all skills used in conversation
- With skill-name: Focuses on specific skill
2. Automatic Reflection
When enabled via /reflect-on, runs automatically at session end via Stop hook.
3. Toggle Commands
/reflect-on- Enable automatic reflection/reflect-off- Disable automatic reflection/reflect-status- Show current configuration
Confidence Levels
HIGH - Explicit corrections:
- User contradicts Claude's approach with specific alternative
- Pattern: "Don't do X, do Y instead"
- Action: Direct updates with deprecation warnings
MEDIUM - Approvals and patterns:
- User approves specific approach
- Pattern succeeds multiple times
- Action: Add to "Best Practices" section
LOW - Observations:
- User questions or suggests alternatives
- Pattern: "Have you considered..." or "Why not try..."
- Action: Add to "Considerations" section
Workflow
- Signal Detection - Scan transcript for corrections/patterns
- Context Analysis - Extract 5-message context around signals
- Skill Mapping - Match signals to relevant skills
- Change Proposal - Generate diff of proposed updates
- User Review - Interactive approval with natural language editing
- Application - Safe YAML/markdown updates with backups
- Git Commit - Automatic commit with descriptive message
Scripts
Core Engine
scripts/reflect.py- Main orchestration logicscripts/extract_signals.py- Pattern detection enginescripts/update_skill.py- Safe skill file updatesscripts/present_review.py- Interactive review interface
Automation
scripts/hook-stop.sh- Stop hook integrationscripts/toggle-on.sh- Enable auto-reflectionscripts/toggle-off.sh- Disable auto-reflectionscripts/toggle-status.sh- Show status
Safety Features
- Timestamped backups before all edits
- YAML validation before writing
- Lock files prevent concurrent runs
- Graceful error handling with rollback
- Git status checks before commits
References
See references/signal-patterns.md for detailed pattern library.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 463k | Updated | Claude CodeCodex | Verified | Install |
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 463k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 318k | Updated | Claude CodeCodex | Unverified | Install |