vulnhunter-security-scanner
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the VulnHunter source code and associated toolsets from Capital One's official GitHub repository (github.com/capitalone/vulnhunter).
- [COMMAND_EXECUTION]: Instructs the user to execute standard development commands including git clone, pip installation of local packages, and the execution of Python scripts for batch scanning and agent operations.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external codebase content, which creates a surface where malicious instructions embedded in source files could attempt to influence the agent's behavior during a scan.
- Ingestion points: Source code files and directories processed by the /vulnhunt and /vulnhunter-fix commands.
- Boundary markers: Relies on a 'Falsification Engine' and 'Capability Filter' to logically verify identified vulnerabilities, though explicit delimiters for prompt interpolation are not detailed in the user guide.
- Capability inventory: The environment allows for file system access, shell command execution via Claude Code, and authenticated network requests to the Anthropic API and GitHub.
- Sanitization: Employs a multi-phase reasoning workflow (Recon, Hunt, Disprove, Filter) to validate findings and minimize the impact of misleading or malicious patterns in the input data.
Audit Metadata