crypto-tweet

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

• Third-party content exposure detected (high risk: 0.90). The skill's Step 1 Research (and the evals) require running open web searches like " crypto news discussion" and "reddit twitter" to ingest headlines and community sentiment from public sites/social media, which the agent then uses to draft tweets and image prompts—exposing it to untrusted third‑party content that can influence actions.