cross-model-review
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: This skill is designed to analyze code changes and diffs which are external, untrusted inputs.
- Ingestion points: The agent reads specific code files and commit ranges as defined in the review brief template (review-brief-template.md).
- Boundary markers: The protocol attempts to mitigate risk by using a separate scratch brief file to isolate the code from the primary session context and providing explicit 'read-only' and 'no sub-agent' rules to the reviewing model.
- Capability inventory: The executing agent has capabilities for running test suites and merging code changes (references to godmode:test-first and godmode:merge-protocol).
- Sanitization: The process relies on model-to-model instructions and a manual reconciliation step rather than automated content filtering to prevent the secondary reviewer from being influenced by malicious content inside the code files.
Audit Metadata