llm-doc-gen

Convert your codebase into navigable documentation by running llm-doc-gen to analyze source files directly and produce organized topic files. The Skill generates architecture overviews, developer guides, and API references without requiring AST parsing, creating sharded documentation under architecture/, guides/, and reference/ directories for easier maintenance and navigation.

Yes, llm-doc-gen maintains state-based resumability via a project-doc-state.json file, allowing you to pause and resume interrupted scans from checkpoints rather than restarting from the beginning.

llm-doc-gen applies to complex applications with multiple modules where in-depth interpretation of implementation details and concept synthesis is required. It works on any codebase structure, scanning source directories to extract architecture patterns, design decisions, and developer guidance through direct source reading.

The Skill reads source code directly without requiring AST parsing, relying on LLM analysis to interpret implementation details and synthesize structured documentation. This approach supports workflow-driven orchestration and multi-agent consultation for flexible codebase understanding.

llm-doc-gen creates sharded documentation—organized topic files instead of monolithic documents—distributed across architecture/, guides/, and reference/ folders. This structure improves navigation, maintenance, and discoverability compared to single-file documentation.

llm-doc-gen depends on cursor-agent, gemini, and codex to function. Ensure these dependencies are available in your workflow before initiating code analysis and documentation generation.