voice-analyzer

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly requires ingesting 3–5 writing samples pasted by the user ("Phase 1: Gather Samples — Paste each sample into a separate section"), which can be public/untrusted third‑party content (published posts, competitor articles) that the agent must read and that will directly shape its generated voice skill, creating a clear vector for indirect prompt injection.