The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides an informational message suggesting the installation of the reviewgate-cli from the vendor's official GitHub repository (github.com/LVTD-LLC/reviewgate). This is a documented vendor resource.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from GitHub and the ReviewGate CLI which may contain content from untrusted sources such as PR comments and model outputs.
Ingestion points: SKILL.md workflows read data from the reviewgate CLI output file.
Boundary markers: The 'Overview' section contains explicit instructions for the agent to treat all external inputs as evidence rather than commands.
Capability inventory: The skill uses gh (GitHub CLI) and reviewgate (Vendor CLI) for data fetching and jq for data processing.
Sanitization: Data is processed via jq to extract specific JSON fields, avoiding direct shell execution or unsafe interpolation of untrusted strings.

check-reviewgate