capturing-learningslisted

# Capturing Learnings ## Overview Every real-world learning — a gotcha in a specific framework, an incident that revealed a missing convention, a pattern that turned out to work well — gets routed to the right durable location. The skill's job is not to write documentation. It is to ask: "Where does this belong so that it helps the right people at the right time?" ## When to Use Run this skill when you encounter: - **A gotcha:** "I wasted 3 hours because I didn't know X" — this should never happen to the next developer. - **A new convention:** The team agreed on a pattern that isn't written down anywhere yet. - **An incident post-mortem:** Root cause identified — the fix is code, but the learning is knowledge. - **A retro output:** "We should always do X" or "never do Y" emerged from a retrospective. - **A workflow improvement:** A step was missing from a skill or reference that would have saved time. - **A false positive in a shared doc:** Something in a team or shared reference doc turned out to be wrong for this project. The trigger phrase: "I wish someone had told me that." ## Decision Tree Given a learning, route it to one of four destinations. Work through the questions in order. ### Question 1: Is this specific to one repo? If the learning only applies to a single codebase and would confuse or mislead developers on other projects: **Route to: Repo's CLAUDE.md** Edit the repo's `CLAUDE.md` directly. Add it to the most specific applicable section. If no sect