last30days

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches public Reddit/X threads and arbitrary web pages (see scripts/lib/openai_reddit.py and scripts/lib/reddit_enrich.py for Reddit fetching, scripts/lib/xai_x.py for X, scripts/lib/websearch.py and the output_result "### WEBSEARCH REQUIRED ###" block plus SKILL.md) and passes those untrusted, user-generated sources into Claude to synthesize findings and decide output formats, so external page content can materially influence the agent's actions.