The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitLab including issue descriptions and merge request bodies. 1. Ingestion points: scripts/gitlab.py (summary formatting functions). 2. Boundary markers: Absent; project data is presented as standard markdown without isolation. 3. Capability inventory: The agent is authorized to use glab for high-impact actions like merging code and creating releases. 4. Sanitization: Absent. This creates a surface where malicious instructions embedded in GitLab content could attempt to influence the agent's behavior. \n- [COMMAND_EXECUTION]: The skill invokes the glab CLI via subprocess.run() in scripts/gitlab.py. The implementation prevents command injection by passing arguments as a list rather than a shell string. Additionally, the script uses argparse to ensure that parameters like issue numbers and result limits are validated as integers. \n- [EXTERNAL_DOWNLOADS]: The skill utilizes the official glab CLI tool and directs users to the official GitLab repository for installation instructions. This is a legitimate dependency for the skill's functionality.

gitlab