github-review-suggestions
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from GitHub Pull Requests, which could be manipulated by contributors to influence agent behavior.
- Ingestion Points: The skill fetches PR metadata, file diffs, and existing comments using the
ghCLI andgitcommands, which are then processed by the agent. - Boundary Markers: The instructions in
SKILL.mdexplicitly warn the agent to treat PR data as untrusted third-party text and not follow instructions embedded within it, providing a soft boundary. - Capability Inventory: The skill is authorized to post comments to GitHub repositories using the
gh apitool and can execute local Node.js/Bun scripts for diff calculations as defined in itsallowed-toolsmetadata. - Sanitization: No programmatic sanitization or structured delimiters are applied to the external PR content before it is interpolated into the agent's context.
Audit Metadata