perspective-swarm

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

• Third-party content exposure detected (high risk: 0.90). Stage 2 explicitly instructs each parallel agent to perform 1–2 WebSearch queries and include supporting evidence/sources in their perspective outputs, and the convergence algorithm and handoff logic use the "research_backed" flag and evidence-weighted scoring to influence synthesis and downstream handoff decisions—meaning untrusted public web content can materially change agent behavior.