article-evaluation-pipeline

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

• Third-party content exposure detected (high risk: 1.00). The skill's Phase 1: FETCH explicitly instructs fetching arbitrary URLs (via WebFetch or curl) and the Example 1 shows evaluating https://example.com/posts/my-article/, meaning the agent will ingest open/public third-party web content and directly use it to drive validation, analysis, and reporting — exposing it to potential indirect prompt injection.