The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill implements destructive commands like glab repo delete and specifically documents the -y, --yes flag to bypass confirmation prompts. In an autonomous agent context, this creates a significant risk of accidental or malicious data loss.
[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from external sources (GitLab project metadata, search results) and possesses high-impact capabilities through the Bash tool. A malicious project could use its name or description to attempt to influence the agent's logic.
Ingestion points: glab repo search, glab repo view, and glab repo contributors output.
Boundary markers: Absent. No specific instructions are provided to the agent to treat external project data as untrusted.
Capability inventory: Full Bash tool access, file read/write through glab operations.
Sanitization: No input validation or sanitization of repository paths or names is mentioned.
[EXTERNAL_DOWNLOADS] (SAFE): While the skill depends on the external glab binary being present on the system, it does not attempt to download or install it at runtime.

gitlab-repo