| name | claudeception |
| description | Continuous learning system. After completing tasks, extract non-obvious discoveries and update repo knowledge (CLAUDE.md) with reusable patterns. |
| metadata | {"version":"1.0.0","type":"post-session"} |
Claudeception — Knowledge Extraction
After completing a task, extract and preserve reusable knowledge.
When to Extract
- Non-obvious solutions: Required significant investigation
- Project-specific patterns: Conventions not documented elsewhere
- Tool integration knowledge: Undocumented API/library behaviors
- Error resolution: Misleading error messages with actual root causes
- Workflow optimizations: Multi-step processes that can be streamlined
Quality Criteria
- Reusable: Will help with future tasks
- Non-trivial: Requires discovery, not just docs lookup
- Specific: Exact trigger conditions and solution described
- Verified: Solution actually worked
Extraction Process
- Identify: What was non-obvious about the solution?
- Document: Update CLAUDE.md with the learning
- Structure: Include problem, trigger conditions, solution, verification
Self-Reflection Prompts
- "What did I learn that wasn't obvious before starting?"
- "If I faced this problem again, what would I wish I knew?"
- "What error message led me here, and what was the actual cause?"