codebase-analyzer

The Codebase Analyzer is a specialized skill designed to eliminate the maintenance burden of documentation by keeping project context files like CLAUDE.md or AGENTS.md continuously up to date. It is intended for software engineers and development teams working in environments where architectural drift, changing build commands, and evolving code conventions often lead to stale documentation. By integrating directly into the Claude Code workflow, the tool provides a 'zero-config' experience that tracks edits, file movements, and bash operations to regenerate memory sections in real-time without manual intervention.

• Automatically scans directory structures and detects frameworks including Node.js, Python, Rust, Go, and container-based builds.

• Employs a robust marker-based system (AUTO-MANAGED tags) to safely update documentation while strictly preserving user-defined MANUAL sections.

• Supports hierarchical documentation for monorepos, allowing for separate subtree CLAUDE.md files in directories like /src, /lib, /packages, and /apps.

• Features an isolated agent-based processing model that performs documentation updates in a separate context window, ensuring that memory maintenance does not consume the primary conversation's token limits.

• Offers interactive initialization via the /auto-memory:init command to analyze existing projects and migrate them to the auto-managed format with user approval.

• To get started, execute /auto-memory:init in your project root to scan your codebase and establish the initial memory configuration.

• Use the /auto-memory:sync command to manually refresh project memory if you have made significant changes outside of Claude Code, such as via IDE refactoring or direct shell modifications.

• Configure advanced triggers in .claude/auto-memory/config.json to switch between default real-time tracking and git-mode, which updates documentation specifically upon commit operations.

• The tool is highly token-efficient, utilizing non-output hooks for file change tracking, ensuring no overhead during standard coding sessions.

• Adhere to the provided structure templates to ensure compatibility with the automated parsing logic; the analyzer strictly adheres to defined marker formats to prevent corruption of your documentation files.