verified-infographic-generator

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

• Third-party content exposure detected (high risk: 0.85). Phase 2 “Source Discovery” performs runtime web searches and ingests quoted sentences from outsider-authored public web pages into Claude’s verification workflow (which then becomes LLM context for Phase 3/4/5).

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

• Potentially malicious external URL detected (high risk: 0.70). The skill explicitly fetches and quotes external web sources at runtime (e.g., https://pubmed.ncbi.nlm.nih.gov/31001682/ and https://www.interface.com/...) in Phase 2 and then copies those quoted passages/numbers verbatim into the Phase 6 image prompt, so remote page content directly controls the prompt text produced by the agent.