apiclaw

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the processing of untrusted data from an external source.
  • Ingestion points: The scripts/apiclaw.py script retrieves raw Amazon review text from the api.apiclaw.io endpoint.
  • Boundary markers: The prompt generated in render_review_map_prompt uses triple-quote delimiters and explicit instructions (e.g., "Respond ONLY with a JSON object", "do not infer") to isolate the untrusted content.
  • Capability inventory: The skill can perform network requests to the author's API domain. No subprocess execution, arbitrary file writes, or persistent mechanisms were found.
  • Sanitization: Review text is truncated to 500 characters before being passed to the LLM to limit the injection surface.
  • [COMMAND_EXECUTION]: The skill executes a local Python utility script (scripts/apiclaw.py) to manage API interactions.
  • The script handles authentication, rate limiting, and request formatting for the vendor's commerce data endpoints. It uses standard libraries for networking and data processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 07:05 PM
Security Audit — agent-trust-hub — apiclaw