auto-learnlisted

# Auto-Learn — The System That Remembers and Fixes Itself When something goes wrong, this skill ensures it never goes wrong the same way again. ## The Loop ### 1. Detect Something failed, broke, or required manual correction: - A command that should not have run - A file committed without reading the diff - A skill that did not activate when it should have - A pattern that keeps repeating Ask: "What went wrong? What was the gap?" ### 2. Diagnose Identify the root cause — not the symptom: - Was it a missing hook? → Create the hook - Was it a missing rule in CLAUDE.md? → Add the rule - Was it a skill with wrong triggers? → Fix via /autoresearch - Was it a missing check? → Add to safety guard or system awareness - Was it knowledge that was lost? → Add to memory with context Ask: "Why did the system allow this to happen?" **The fix does not live where the bug manifests — it lives at the node where the relational condition was missing.** A patch at the point of failure is accumulation; an inversion at the node is maturation. If the same error reappears in a different context, the patch was applied to the symptom, not to the node. Trace back through the chain of successive steps until you find where the missing condition sits. That is where the fix belongs. A separate observer (not the producer of the error) analyzes the path — not the result. Looking at the output tells you what failed. Looking at the trajectory tells you where the system could have caught it earlier. ###