adversarial-reviewer
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is entirely prompt-driven and does not require any external tools or dependencies, which minimizes the attack surface.
- [SAFE]: Command execution is limited to local git operations (git diff, git diff --cached) for gathering context, which is standard and expected behavior for a development-focused code review skill.
- [SAFE]: The inclusion of a Security Auditor persona actively encourages the detection of vulnerabilities such as hardcoded secrets, injection flaws, and broken access control in the reviewed code, promoting secure coding practices.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of code and git diffs which could contain adversarial instructions meant to influence the reviewer's output. Ingestion points: Git diff outputs and full file contents read from the local filesystem. Boundary markers: The skill does not define specific delimiters to separate instructions from the data being reviewed. Capability inventory: The skill has read-only access to the filesystem; it lacks network access or the ability to execute the code it analyzes. Sanitization: No explicit sanitization or filtering of the input code is performed.
Audit Metadata