deep-research

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs web searches of public community and code sites (e.g., "VCV Rack community forum search via WebSearch" and GitHub issue queries in deep-research Level 2) and the market-research-reports flow requires using the research-lookup scripts to gather public market data, and those external, user-generated sources are synthesized and used to produce recommendations and drive follow-up actions (e.g., "Apply recommended solution" → invoke module-improve), so untrusted third-party content can influence agent decisions.