What problem does it solve?
Long-running AI collaboration loses context: agents forget user preferences, repeat mistakes, and re-explore project structure every session. This Skill maintains a durable knowledge layer so stable facts, decisions, and corrections persist across conversations.
Core Features & Use Cases
- Fact Routing: Directs each piece of knowledge to the right location—project root AGENTS.md for architecture and commands, workspace AGENTS.md for environment facts, memory/ topic files for user preferences and lessons, or Skills for repeatable procedures.
- Memory Organization: Maintains MEMORY.md as a topic index, splits overloaded topic files, merges duplicate conclusions, and preserves time context on time-sensitive facts.
- Conservative Write Policy: Writes only stable, verified increments directly; asks for confirmation before deletions, large overwrites, uncertain inferences, or sensitive personal data.
- Use Case: After a user corrects the agent twice about preferring pnpm over npm, the Skill records the preference in the workspace memory collaboration topic and updates the project AGENTS.md install command, so future sessions apply it automatically.
Quick Start
Ask the agent to remember a stable preference or project fact, for example: remember that this project uses pnpm and always run tests with pnpm test before committing.
