The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses basic shell commands (ls, grep, head, cat, diff, which, test) to inspect the repository for languages, frameworks, and build targets. These operations are limited to local filesystem analysis and do not involve executing external scripts or non-standard tools.
[PROMPT_INJECTION]: The skill ingests content from repository files such as README.md, Makefile, and source code to generate the CLAUDE.md file. While this involves processing untrusted local data, the skill follows a structured pipeline with explicit validation gates and focuses on metadata extraction, which mitigates the risk of indirect instruction override.
[SAFE]: The skill performs legitimate repository documentation tasks using least-privilege principles, avoids network connectivity, and implements safety measures such as writing to temporary files (CLAUDE.md.generated) when a target file already exists.

generate-claudemd