The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary orchestration script, scripts/security_check.py, utilizes subprocess.run to execute security tools (gitleaks, trivy, semgrep) and git commands. These calls use argument lists rather than shell strings, which is a secure practice to prevent shell injection. The script also provides a mechanism for users to pass extra arguments via the SECURITY_CHECK_ARGS environment variable, which is safely parsed using shlex.split and argparse.
[EXTERNAL_DOWNLOADS]: The skill recommends installing security utilities from official and well-known package managers such as PyPI, Homebrew, and Winget. It explicitly advises against using unverified shell scripts (curl-pipe-bash) for tool installation, favoring official documentation links for manual setup if a package is unavailable.
[INDIRECT_PROMPT_INJECTION]: As a security scanning utility, the skill inherently possesses an attack surface related to processing untrusted repository code. However, it mitigates risk by using local scanners with pre-defined rule sets (security/semgrep-rules.yml) and avoids the dynamic interpolation of untrusted file content into high-privilege agent instructions.
[PROMPT_INJECTION]: Analysis of the instructional content in SKILL.md and docs/README.md shows no attempts to override agent behavior, reveal system prompts, or bypass safety filters. The language used is purely functional and instructional.

security-setup