superreview
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill directs the agent to execute existing project-defined scripts for building, linting, and testing to validate code changes. Evidence in
references/review-contract.mdspecifies running these commands in an isolated workspace and recording the exit status and output. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface by ingesting and acting upon instructions and rules found within the project being reviewed, such as those in
AGENTS.mdand.codex/rules/. - Ingestion points: Target repository files, Git status and diff outputs,
AGENTS.md, and project-specific rule files. - Boundary markers: Absent; the instructions do not require the use of delimiters or explicit 'ignore' warnings when processing untrusted code or documentation.
- Capability inventory: Local file system access, execution of Git commands, execution of local build and test scripts, and the ability to write report files to the filesystem.
- Sanitization: Absent; the agent is instructed to read and follow project-level rules as authoritative guidelines for its review process.
Audit Metadata