The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied community briefs and feedback. It mitigates potential risks through instructions that require the agent to treat such data as "unverified" or "supplied," and to explicitly avoid inventing safety-critical procedures based on incomplete or unauthoritative input.
[COMMAND_EXECUTION]: The skill includes a Python utility (scripts/validate_guide.py) used to validate the structure of community plans. Static analysis identified subprocess.run calls in the associated test suite (scripts/test_validate_guide.py), which are used correctly and safely to verify the validator's command-line interface within a controlled testing environment.
[SAFE]: The Python validation script incorporates security best practices by explicitly checking for and rejecting absolute paths, parent directory traversal components (..), and symlinks that resolve outside the intended directory. These checks effectively prevent path traversal vulnerabilities during plan validation.

community-guides