Skip to content

Latest commit

 

History

History
44 lines (28 loc) · 1.64 KB

File metadata and controls

44 lines (28 loc) · 1.64 KB

Using the codex-keeper

codex-keeper creates AI-focused documentation for your codebase, structured for navigation by subagents.

What it produces

codex-keeper writes to .ai-docs/ and updates your project's CLAUDE.md with a ## Generated Documentation section so other subagents know where to find the docs.

Output Purpose
.ai-docs/DOCUMENTATION_MAP.md Coverage index with staleness tracking and priority queue
.ai-docs/reference/*.md Architecture, type system, store maps, etc.

For complex codebases it generates a features/ subfolder with docs by feature. Once they exist, you and your subagents can reference them by name:

@web-developer see X feature documentation to understand how X works, then implement Y

Setting up a documentation repository

The recommended setup keeps .ai-docs/ as a standalone repo symlinked into your project. This lets you version documentation independently, share it across team members, and separate doc history from code history.

Keeping docs fresh

Docs drift quickly. Run codex-keeper weekly to catch areas that have changed.

Modes

New documentation: Creates docs for undocumented areas. Start here for a fresh codebase.

@codex-keeper initialize documentation for this codebase

Update: Refreshes all or specific docs

@codex-keeper update any docs that are stale
@codex-keeper the batch processing feature changed significantly, update its docs