claygent

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime workflow is to “Visit /company_domain” and extract answers from web pages the agent browses, which are outsider-authored free text from public websites that the LLM ingests for extraction.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly instructs the agent at runtime to "Visit /company_domain" (i.e., fetch arbitrary company websites) and use that site content to answer prompts, so external website content is fetched during runtime and directly controls the agent's outputs.