The Agent Skills Directory

SAFE (INFO): The skill contains no executable code, binary files, or active scripts. All 25 files are markdown-based documentation or metadata. No malicious patterns such as obfuscated code, credential harvesting, or prompt injection were identified.
COMMAND_EXECUTION (LOW): The documentation describes various CLI commands (e.g., npx undocs dev, npx undocs build) and package manager operations. These are standard development instructions for the tool being documented and do not constitute a malicious attempt to hijack the agent's environment.
EXTERNAL_DOWNLOADS (LOW): The documentation references external tools like undocs, giget, and nuxt. While these are not in the predefined 'Trusted Sources' list, they are well-known open-source projects within the UnJS/Nuxt ecosystem. The references are informational and standard for technical manuals.
DATA_EXFILTRATION (INFO): The skill mentions endpoints like /raw/*.md and /llms.txt, which are designed to expose documentation content to LLMs and CLI tools (via curl or Accept headers). These are documented features for documentation accessibility and do not involve sensitive user data exposure.

undocs