reviewing-with-peer-agents
Pass
Audited by Gen Agent Trust Hub on Aug 25, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to handle and process data from external AI models (peer agents), which is an inherent attack surface. However, the instructions incorporate strong defensive patterns: it mandates that every response be treated as untrusted review input, requires verification of material claims against the source project, and establishes a strict 'resolve-authorized' mode that prevents unauthorized file modifications, commits, or pushes without explicit user intervention.
- [DATA_EXPOSURE_&_EXFILTRATION]: The skill explicitly prohibits the transmission of private chain-of-thought or unrelated sensitive context to peer agents, limiting data sharing to the minimum sufficient artifacts required for the review.
- [COMMAND_EXECUTION]: The skill does not define or execute any shell commands, tools, or APIs directly. It functions as a procedural guideline for existing agent capabilities.
Audit Metadata