community-signal-digest

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly calls search_external_content to ingest public, user-generated posts from DEV.to, Hacker News, YouTube, and Hashnode (and optionally Common Room sources like Reddit, Stack Overflow, X, Slack, Discord, Discourse) and then reads and interprets those items to classify, prioritize, and automatically draft responses (see Steps 4, 5, 9, and 11), so untrusted third‑party content can materially influence agent actions.