sentiment-analyst

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly ingests public social media content—e.g., derivatives_sentiment(action="reddit_trending") returning top mentions from crypto subreddits—and instructs the agent to use that community discussion data in the full deep-dive synthesis (see SKILL.md and references/signal-guide.md), so untrusted, user-generated third‑party content can influence analysis and actions.