Patpat Learn
When invoked directly, read the operating protocol in full. Do not load the router.
Read encode lessons, repository truth, and apply the learning playbook.
Use evidence from the active task: user corrections, failed verification, repeated retries, review findings, or verifier defects. Separate one-off facts from recurring failure modes. Encode only the latter.
Choose the strongest narrow mechanism that prevents recurrence: type or contract, focused test, lint or validator rule, deterministic script, workflow instruction, then documentation. Prefer enforcement over reminders and update an existing authoritative location before creating a new file. Mine this conversation for recurring working-style or failure rules. Propose edits to existing files only. Present the proposal and wait for approval. Do not auto-apply. There is no new SKILL.md and no *-mode mint on this path. Never create a new SKILL.md. Never mint a personal *-mode skill.
Do not mine unrelated conversations or store sensitive task data. Do not modify shared or user-global rules without explicit authority. Verify that the chosen mechanism detects or prevents the original failure, then report what remains unprotected.
Proof closure
Close repository mutations through: