What problem does it solve? Valuable insights discovered during a work session — invariants, pitfalls, architectural decisions — are lost when the session ends. This Skill captures that hard-won knowledge and stores it in a persistent memory provider so future sessions and agents can reuse it. ## Core Features & Use Cases - Knowledge Extraction: Scans conversation history, tool outputs, and touched files to identify reusable, non-obvious facts that clear a strict write bar. - Dedupe and Structure: Checks existing memories before writing, edits overlapping entries, and organizes knowledge under stable slash-separated topic paths like pitfalls/<topic> or decisions/<topic>. - Provider-Based Persistence: Hands each write to the active memory-provider driver defined in <project-root>/docs/memory-providers.md, keeping storage mechanics decoupled. - Use Case: After a long debugging session that uncovered a non-obvious API rate-limit behavior, run this Skill to distill that finding into a memory entry so the next agent never rediscovers it the hard way. ## Quick Start Ask the agent to save what we learned this session to memory before wrapping up the task.
