agentic-research

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

• Third-party content exposure detected (high risk: 1.00). The skill's playbook explicitly requires agents to discover, crawl, scrape, and relay arbitrary web URLs (see SKILL.md Team Structure and "URL Report"/"Source Relay" formats) and explicitly lists community/user-generated sources like Reddit and Stack Overflow in references/source-quality-tiers.md and references/message-templates.md, meaning the agents will ingest and act on untrusted third‑party content that could carry indirect prompt injections.