agentic-voice-workflow

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

• Third-party content exposure detected (high risk: 0.80). The workflow explicitly lists "External Dependencies: Web search" in Phase 0 and references a "deep_research" reference agent whose pattern is "Web research → LLM synthesis → markdown report" (SKILL.md and references/full-workflow.md), indicating the agent workflow consumes open/public web content that could carry untrusted, user-generated instructions.