What problem does it solve? Corrections, preferences, and lessons from a conversation are lost when the session ends, forcing users to repeat the same instructions. This Skill captures that feedback and turns it into persistent global assets (rules, skills, hooks, memory) so it applies in future sessions. ## Core Features & Use Cases - One-off vs. recurring judgment: First decides whether feedback is a one-time instruction or a recurring convention, avoiding asset pollution from single-use requests. - Placement decision: Routes each learning to the right location (rules/, skills/, settings.json hooks, CLAUDE.md, or project memory) and to public (home/) vs. private/ storage based on sensitivity. - Approval and maintenance workflow: Requires AskUserQuestion confirmation before editing, then performs inventory cleanup (split, merge, delete, cross-reference) and commits changes to the dotfiles repository. - Use Case: After a user corrects the assistant twice about commit message style, invoke this Skill to propose a new rule file, get approval, place it in the correct directory, cross-link related rules, and commit the change. ## Quick Start Ask the assistant to run /learn-rules to turn the recent corrections from this conversation into a persistent rule or skill.
