wos-navigate-pages
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external search results from the Web of Science API, creating a surface where malicious instructions could be embedded in data such as paper titles or abstracts.
- Ingestion points: The skill fetches data from the WoS API at /api/wosnx/core/runQuerySearch.
- Boundary markers: No delimiters or specific instructions are provided to the agent to treat the retrieved content as data rather than instructions.
- Capability inventory: The skill utilizes evaluate_script to run JavaScript that performs network requests and data processing.
- Sanitization: No sanitization or filtering is applied to the record fields returned from the API.
- [DYNAMIC_EXECUTION]: The skill provides JavaScript code templates containing placeholders that are populated with search parameters and executed within the browser environment at runtime.
- [DATA_EXFILTRATION]: The skill includes logic to extract a Session ID (SID) from browser performance resource entries. This involves accessing sensitive session-level information to facilitate authenticated API requests.
Audit Metadata