marketing-launch

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly instructs the agent to fetch and interpret public, user-generated web content—e.g., Research Mode's agent-browser commands to open Product Hunt, LinkedIn, Reddit, competitor blogs and Context C's live-URL audit—and to use those findings to drive launch decisions and actions, which creates a clear vector for indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's Research Mode explicitly runs agent-browser to open and "get text body" from external pages (e.g., https://www.producthunt.com/topics/{category}), which at runtime fetches remote page content that is injected into the agent's context to drive extraction and recommendations, so this is a runtime external dependency that can directly control prompts.