codebase-onboarding-generator

To generate a CLAUDE.md for an unfamiliar codebase, run a local read-only analyzer on the repository path to detect project type, directory structure, commands, and conventions, then synthesize the results into the documentation file.

Yes, you can detect project frameworks and coding conventions by analyzing repository files and sampling source code to infer configurations from common build files like pyproject.toml and package.json.

The best way to extract build and test commands is by parsing common build configuration files such as Makefiles, package.json, and pyproject.toml to identify install, build, test, lint, and run commands.

To analyze directory structure while excluding noise, apply a read-only file analyzer that detects meaningful directories like src and tests while explicitly filtering out node_modules and virtual environments.

No, codebase onboarding documentation generation does not require an internet connection because the analyzer operates locally and deterministically on the target repository files.

You should regenerate project documentation after major codebase changes when new frameworks, environment variables, or structural directories are introduced and need to be reflected in the CLAUDE.md file.