review-accuracy-calibration
Pass
Audited by Gen Agent Trust Hub on Apr 8, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill contains purely instructional content and decision-making logic for code review tasks.
- [PROMPT_INJECTION]: No attempts to override agent safety guidelines or system instructions were found. The instructional language is aligned with the stated purpose of improving review quality.
- [DATA_EXFILTRATION]: There are no network operations, external requests, or access to sensitive local files such as credentials or SSH keys.
- [COMMAND_EXECUTION]: No shell commands, system calls, or subprocess executions are defined within the skill content.
- [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts or packages from remote sources.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted code as its primary data source. While this represents a standard attack surface for code review tools, the guidelines explicitly include verification steps (checking framework conventions, reading surrounding context, and requiring measurable impact) which serve as mitigations against being misled by malicious patterns in the code under review.
Audit Metadata