biz-scan
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing untrusted codebase data.\n
- Ingestion points: Source code files are read and searched using the
ReadandGreptools.\n - Boundary markers: No specific delimiters or safety instructions are provided to the agent to ignore instructions embedded within the codebase content.\n
- Capability inventory: The skill's permissions are restricted to
Read,Grep, andGlob; it has no ability to write files, execute arbitrary code, or perform network operations.\n - Sanitization: Content from the codebase is processed without prior validation or escaping.\n- [COMMAND_EXECUTION]: The skill executes shell-based search commands (
grep) to locate data models, API endpoints, and tracking events. These commands are static and scoped to the project directory, representing standard diagnostic behavior for a business intelligence agent.
Audit Metadata