signal-detection-pipeline

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs running public, user-generated sources—e.g., reddit-post-finder (Reddit posts), linkedin-post-research/linkedin-commenter-extractor (LinkedIn content), luma-event-attendees (event URLs) and web search for enrichment—so the agent ingests and acts on untrusted third-party content that can influence lead scoring and outreach decisions.