zapier-demo

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

• Third-party content exposure detected (high risk: 0.85). Step 2 requires fetching Zapier’s marketing page at https://zapier.com/apps/{app-slug}.md (public web content), and that fetched page’s readable text is ingested into the agent to extract available actions—an outsider-authored free-text source.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs the agent at runtime to fetch the app marketing page (e.g., https://zapier.com/apps/{app-slug}.md) and to use that fetched content to determine available actions and pick the demo prompt, so external content directly controls the agent's prompts.