streak

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly performs optional external "research" (e.g., "scan for relevant news/updates" and "search for updates relevant to stack/topic" in Flow 2 Step 3.5 and the BUILDING type notes) and also accepts linked "Reference Links" in context.md, which means the agent may fetch and summarize open/public third‑party content (news, websites, libraries) as part of its workflow.