drugbank-search
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on local script execution (
scripts/search) for searching and initial configuration. - Evidence: The skill provides instructions for the agent to locate the script in the plugin cache and execute it with user-provided queries or setup parameters.
- [INDIRECT_PROMPT_INJECTION]: The skill processes structured data (JSON) retrieved from the Valyu API, which contains drug information, mechanisms, and interactions from external sources.
- Ingestion points: Results are returned from the
scripts/searchcommand in JSON format. - Boundary markers: None explicitly defined in the provided output format, though the data is structured.
- Capability inventory: The agent reads and processes the returned JSON data to provide information to the user.
- Sanitization: Not explicitly implemented in the prompt instructions; the agent is expected to parse the JSON fields (
content,title, etc.) directly. This represents a standard surface area for indirect injection from the source database, though the risk is minimal for this specific use case.
Audit Metadata