The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses gh api and git commands to research candidate profiles and activity. These commands are constrained by the allowed-tools frontmatter to specific subcommands (gh api:* and git:*), which reduces the risk of arbitrary command injection.
[PROMPT_INJECTION]: The skill processes untrusted data from external sources (GitHub PR descriptions, repository code, and issue comments). This creates a surface for indirect prompt injection if a candidate includes malicious instructions in their public GitHub content.
Ingestion points: External data is ingested via gh api search/issues and gh api search/code (SKILL.md).
Boundary markers: The instructions do not define clear delimiters or warnings to ignore instructions embedded in the analyzed GitHub content.
Capability inventory: The skill possesses Write and Edit capabilities, intended for updating CONTRIBUTORS.md (SKILL.md).
Sanitization: There is no explicit sanitization or filtering logic mentioned for the content retrieved from GitHub API calls.

candidate-evaluation