What problem does it solve? AI agents lose hard-won project knowledge between sessions, repeating mistakes and rediscovering the same conventions. This Skill provides a persistent, project-scoped memory store under .rc/memory/ so decisions, corrections, and verified workflows survive across turns and sessions. ## Core Features & Use Cases - Facts: Records durable decisions, conventions, gotchas, and glossary terms as one markdown file per fact, indexed in INDEX.md for fast recall before planning or implementing. - Lessons: Captures what failures and user corrections taught via scripts/lessons.mjs, with recurrence tracking, candidate-to-confirmed promotion, quarantine, and a rendered LESSONS.md playbook. - Golden paths: Promotes verified multi-step routes (commands, required order, dead-ends) into project-local skills once they pass a three-part gate: a passing check, a named failure pattern, and a ruled-out dead-end. - Use Case: After a user corrects your approach to database migrations, record a grounded lesson with --signal user_correction; in the next session, confirmed lessons load automatically as design constraints before planning. ## Quick Start Ask the agent to check project memory for anything relevant to the current task before starting work, and to record what it learns afterward.
