article

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

• Third-party content exposure detected (medium risk: 0.30). The workflow reads external GitHub repositories, issues, PRs, and web pages via WebSearch, WebFetch, and gh api, allowing outsider-submitted text into the agent's ingestion path.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (medium risk: 0.30). The skill instructs fetching content from external APIs (Semantic Scholar and Replicate) via curl during runtime execution.