diagnosing-dependabot-alerts

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

• Third-party content exposure detected (high risk: 0.85). The workflow has the agent call gh api ... dependabot/alerts/<N> and then uses the returned alert fields (including summary and vulnerability range text) directly in the agent’s reasoning context, which is outsider-authored free text from GitHub Dependabot/GHSA at runtime.