marketing-report-expert

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

• Third-party content exposure detected (high risk: 0.85). Phase 1 “Brand Cloning” uses browser_subagent to visit a user-provided URL at runtime and extract logo/colors/fonts from the fetched public webpage text, which is outsider-authored free text.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs the agent to crawl a user-provided website at runtime (e.g., https://apple.com) using browser_subagent and extract logo/colors/fonts that are injected into downstream prompts and passed to the PDF renderer, so fetched remote content directly controls agent instructions/execution.