peer-review
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted Material (code artifacts, text) and includes it in prompts sent to the Codex agent, creating a surface for indirect prompt injection. Ingestion points: Material identified in Step 1. Boundary markers: Explicit instructions to pin the review to the diff and treat other topics as off-topic. Capability inventory: Tool invocation of /codex-exec and file writing to .turbo/. Sanitization: Codex is instructed to bound its reads to the material and criteria.
- [COMMAND_EXECUTION]: The skill performs local file system operations by writing git diffs or working-tree scopes to local files within the .turbo/ directory during the preparation step.
Audit Metadata