orchestrating-vulnerability-research
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates the analysis of untrusted data (source code, binaries, web responses), which creates a surface for indirect prompt injection attacks. Malicious content within the analyzed targets could potentially influence the behavior of the hunter or critic agents.
- Ingestion points: The workflow ingests target codebases, compiled binaries, and live HTTP request/response data (SKILL.md).
- Boundary markers: The instructions do not specify the use of boundary markers or 'ignore embedded instruction' warnings when interpolating target data into agent prompts.
- Capability inventory: The skill dispatches hunters with capabilities defined in referenced skills (e.g., auditing-code-for-vulnerabilities, testing-web-applications), which typically include file system access and network operations.
- Sanitization: No explicit sanitization, validation, or filtering of the external content is described before it is processed by the agent roles.
Audit Metadata