parallel-ai-search
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted data from the live web through its search and extraction functions, making it susceptible to indirect prompt injection where malicious instructions on web pages could influence the agent's behavior. ● Ingestion points:
scripts/parallel-search.mjs,scripts/parallel-extract.mjs, andscripts/parallel-search-extract.mjs(via Parallel AI API). ● Boundary markers: None; fetched content is returned directly to the agent context as text or Markdown. ● Capability inventory:scripts/lib.mjsperforms network operations via the fetch API and can read local files usingfs.readFile(triggered by the --request flag). ● Sanitization: No sanitization or safety filtering is performed on the extracted content. - [EXTERNAL_DOWNLOADS]: The skill communicates with the Parallel AI API at
https://api.parallel.aito perform its core functions of web searching and data extraction. This is documented and expected behavior.
Audit Metadata