web-utilities-research

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow exposes an outsider prompt-injection surface because users can supply arbitrary URLs/text to the required endpoints /web/scrape and /extract, which fetch and ingest the contents of those outsider-authored web pages at runtime via the Crawlora API.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill calls the Crawlora API at runtime (https://api.crawlora.net/api/v1 and specific endpoints such as https://api.crawlora.net/api/v1/geocoding/reverse?lat=37.4224&lon=-122.0842) to fetch arbitrary webpage content which is returned and typically injected into the agent's context, so remote content can directly influence model prompts.