Verified:GitHub

librarian

Maintain a markdown knowledge base of codebase architecture and decisions.

2|1|Updated Apr 5, 2026

Search Tags:#knowledge management #context retrieval #codebase #verification #architecture documentation #design decisions #refactor maintenance

Ready to run

npx skills add https://github.com/gokuito/librarian --skill librarian-gokuito

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: librarian
Source: https://github.com/gokuito/librarian/tree/main
Command: npx skills add https://github.com/gokuito/librarian --skill librarian-gokuito

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill prevents your AI and your team from repeatedly rediscovering how your codebase works by maintaining an always-current knowledge base.

Core Features & Use Cases

Auto-retrieval before work: Reads .knowledge/_index.md and selectively loads the most relevant architecture/decision/docs files before non-trivial tasks.
After-action documentation: Records design decisions, gotchas, and non-obvious patterns after tasks that change architecture or add important context.
Verification and refreshing: Checks for stale/dead paths and factual mismatches against the current codebase, and updates existing entries after refactors.

Quick Start

Tell Claude Code to bootstrap the knowledge base by running bootstrap the knowledge base.

Frequently Asked Questions about librarian

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I maintain a living codebase knowledge base for AI context retrieval?▼

To maintain a living codebase knowledge base, use a Skill that auto-retrieves relevant architecture files before tasks and records design decisions after changes. It indexes markdown files using YAML frontmatter to provide faster, accurate AI context.

What is the best way to document architecture decisions after refactoring code?▼

The best way to document architecture decisions after refactoring is to record non-obvious patterns and gotchas in a markdown knowledge base. The Skill verifies factual mismatches against the current repository and updates existing entries to keep knowledge accurate.

How does auto-retrieval of codebase architecture work before non-trivial tasks?▼

Auto-retrieval works by reading a central index file and selectively loading the most relevant architecture, decision, and documentation files before starting non-trivial tasks. This ensures the AI has the necessary context without manual searching.

Do I need YAML frontmatter to index my codebase knowledge base?▼

Yes, you need YAML frontmatter to index your codebase knowledge base. The Skill requires YAML frontmatter-driven indexing to selectively retrieve and load the most relevant architecture and decision files for AI context.

Can I use this knowledge management Skill for onboarding new developers?▼

Yes, you can use this knowledge management Skill for onboarding. It applies to onboarding by answering how and why questions about the codebase architecture, allowing new developers to quickly access verified design decisions and documentation.

When should I not use a markdown knowledge base for codebase documentation?▼

You should not use a markdown knowledge base for codebase documentation when changes have trivial architectural impact. The Skill uses controlled recording to ensure only non-trivial design decisions and architectural changes are documented.