suede-code
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is a defensive security tool intended to identify vulnerabilities in codebases. It contains extensive instructions for detecting hardcoded credentials, injection attacks, and authorization failures.
- [COMMAND_EXECUTION]: The skill instructions specify the use of local project tools such as typecheckers, linters, and test suites. These are standard operations for a code review agent and are scoped to the project environment to provide verification of the reviewed code.
- [INDIRECT_PROMPT_INJECTION]: As a code review tool, the skill naturally processes untrusted data in the form of code diffs and file contents. This constitutes a potential attack surface where a malicious file could attempt to influence the reviewer's output.
- Ingestion points: The skill reads 'current source, current diff, local docs' as defined in the 'Operating Stance' (SKILL.md).
- Boundary markers: The skill employs a structured 'Review Contract' and a strictly defined 'Output Shape' to maintain context and intent.
- Capability inventory: The skill has the capability to execute shell commands (linters/tests) and write files (fix mode), which are necessary for its primary function.
- Sanitization: The instructions emphasize 'evidence-based' reporting and 'honestly' reporting results to mitigate deceptive inputs.
Audit Metadata