Agent Docs

A persistent knowledge base for AI coding agents. The agent reads LIBRARIAN.md at the start of each conversation, discovers topics via directory listing, and reads relevant files before working.

Setup

1. Clone to your home directory:

   git clone https://github.com/afspear/agent-docs.git ~/.agent-docs

2. Add this instruction to your AI agent's system prompt:

   You are a librarian of a central knowledge base at ~/.agent-docs/. At the START of every conversation, read ~/.agent-docs/LIBRARIAN.md, list the directory to see available topics, then read any files relevant to the user's task or workspace context. Apply this knowledge throughout the conversation. When the user wants to remember, save, note, or document something, store it in ~/.agent-docs/ following the guidance in LIBRARIAN.md.
   

3. Start using it. Say things like:

   • "Remember that we use Tailwind for styling"
   • "Note: always run tests before pushing"
   • "Save this deployment process"

The agent will create topic files as you work. Your knowledge base grows organically.

How It Works

• LIBRARIAN.md — Instructions for the agent on how to organize and retrieve knowledge
• *.md — Topic files created by the agent (gitignored, yours are private)

The agent uses directory listing to discover topics — no index file to maintain.

License

MIT