learn-evallisted

# learn-eval — Extract, Evaluate, then Save Analyze the current session for reusable patterns, apply a quality gate, determine the right save location (Global vs Project), then write the skill file after user confirmation. ## What to Extract Scan the session for: 1. **Error Resolution Patterns** — root cause + fix + why it's reusable 2. **Debugging Techniques** — non-obvious diagnostic steps, tool combinations 3. **Workarounds** — library quirks, API limitations, version-specific fixes 4. **Project-Specific Patterns** — conventions, architecture decisions, integration patterns Do not extract: typos, one-off syntax errors, or issues caused by a specific outage. ## Process ### Step 1: Review the Session Read the session conversation and identify the most valuable/reusable insight. Focus on insights that would save time in a future session, not on what was done. ### Step 2: Draft the Skill File Use this format: ```markdown --- name: pattern-name description: "Under 130 characters — what this pattern solves" user-invocable: false origin: auto-extracted --- # [Descriptive Pattern Name] **Extracted:** [Date] **Context:** [When this applies] ## Problem [Specific problem this solves] ## Solution [Pattern/technique/workaround — with code examples] ## When to Use [Trigger conditions for future sessions] ``` ### Step 3: Quality Gate Execute **all** checklist items before proceeding: - [ ] `grep ~/.claude/skills/` and `~/.claude/rules/` and relevant project `.claude/sk