biomedical-search
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface.\n
- Ingestion points: External biomedical search results from the Valyu API (PubMed, preprints, clinical trials, FDA labels) enter the agent context through functionality described in
SKILL.md.\n - Boundary markers: Absent; no specific delimiters or instructions to ignore embedded commands within results are provided.\n
- Capability inventory: The skill utilizes the
scripts/searchbash wrapper andscripts/search.mjsNode.js script to perform queries.\n - Sanitization: Absent; no validation or filtering of the retrieved biomedical content is mentioned before it is returned to the agent.\n- [COMMAND_EXECUTION]: Local script execution.\n
- The skill instructs the agent to execute
scripts/searchandscripts/search.mjsusing the Bash and Node.js environments.\n - It utilizes a dynamic shell command (
find ~/.claude/plugins/cache...) to resolve the script's installation path in the user's home directory.\n- [EXTERNAL_DOWNLOADS]: External API communication.\n - The skill makes network requests to
https://api.valyu.ai/v1to retrieve biomedical information.\n - The documentation references external packages
valyu(Python) andvalyu-js(Node.js) for third-party integration purposes.
Audit Metadata