security-audit
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions prioritize security and privacy by enforcing read-only analysis and the protection of sensitive information. The agent is explicitly directed not to disclose discovered secrets, avoid testing production systems, and use only safe verification methods such as source code inspection or repository-native static analysis tools.
- [PROMPT_INJECTION]: The skill processes untrusted data from repositories and pull requests, which is an inherent surface for indirect prompt injection.
- Ingestion points: Repository source code, configuration files, pull request diffs, and dependency metadata as described in the Discovery Workflow.
- Boundary markers: No specific delimiters or markers are defined in the instructions to separate agent instructions from the code being audited.
- Capability inventory: The agent uses shell commands (e.g.,
git status), file system search, and platform-native tools for analysis and verification. - Sanitization: The provided report format strictly prohibits the disclosure of raw secret values, tokens, or personal data, requiring the agent to refer to them only by path or type.
Audit Metadata