What problem does it solve?
It helps you capture a tricky, codebase-specific lesson from a conversation—so future Claude sessions can apply the same learned principle and workflow instead of repeating the same investigation.
Core Features & Use Cases
- Skill extraction with guardrails: Only records insights that are non-googleable, specific to the project, and precise enough to be actionable.
- Separate learnings into Expertise vs Workflow: Stores domain principles (updatable) apart from stable procedures (kept steady).
- Built-in validation: Rejects vague, boilerplate, or overly generic “best practices” to prevent low-quality skills from being saved.
- Project-level learned skill saving: Creates a version-controlled learned skill in the repository under .omc/skills for consistent reuse.
Quick Start
Ask the AI to extract a new learned skill from your latest debugging session, and include the exact error message, relevant file paths with line numbers, the final fix, and the trigger phrases you expect to see next time.