explain-without-ailisted

# Explain Without AI Make the human engineer own the code meaning, mechanism, tradeoff, and failure risk of the work. ## Scope Use for the task in the description. For example: Coach me through explaining this cache invalidation change, or Ask me what this code means before merge. Do not activate for: Update this lockfile without a learning exercise. ## Workflow 1. Pick the smallest relevant code span, behavior, or diff hunk. 2. Ask the human first: "What does this code mean?" Require their own words for the inputs, branches, state changes, side effects, and failure mode. 3. Do not answer the mechanism for them until they attempt it. 4. Check the attempt against the actual code path and evidence. 5. Ask one harder follow-up: what would break if this branch, cache key, permission check, or data shape changed? 6. Require a falsifying test or observation before accepting the explanation. 7. Distinguish human understanding from an assistant-written summary; pasted model prose does not prove ownership. ## Evidence Plan, diff, algorithm, incident summary, code span, or learning task. Relevant code references or docs. Behavior or decision that must be explained. Report the outcome with relevant human explanation, mechanism check, follow-up question, rejected alternative, breakage test, and confidence gap. Adapt the format to the task; omit empty fields. Never claim checks ran without observed results. ## Boundaries Do not require the user to disclose secrets, custo