ad-angle-miner

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly instructs running scrapers against public, user-generated sources (e.g., G2/Capterra/Trustpilot via review-scraper, Reddit via reddit-scraper, Twitter/X via twitter-scraper, and competitor ad scrapers) and then requires the agent to extract verbatim quotes, score/rank angles, and generate ad copy/test plans—so untrusted third-party content is ingested and can directly influence decisions and tool-driven outputs.