What problem does it solve? AI agents repeat the same mistakes across conversations because lessons learned in one session are lost when it ends. This Skill persists error corrections, task workflows, codebase conventions, and domain terminology into structured files so future sessions start with accumulated knowledge. ## Core Features & Use Cases - Trigger-Based Capture: Activates after error recovery, when users share domain knowledge, or when existing captured learning becomes outdated. - Typed Learning Files: Classifies lessons into task workflows (task-<name>.md), repository/service conventions (convention-<name>.md), and a shared terminology.md, stored under ~/.agents/<workdir>/captured-learning/. - Memory Index Maintenance: Updates a Captured Learning section in AGENTS.md, CLAUDE.local.md, or MEMORY.md so relevant lessons are discoverable at session start. - Use Case: After a user corrects your log-search approach twice, capture the correct procedure as task-log-search.md with the error context, so the next session applies the right method immediately. ## Quick Start After I correct a mistake or share a team convention, capture the lesson into a persistent learning file and update the memory index.
