ai-eval-ci

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime invokes external packages that are fetched-and-executed (e.g., "npx promptfoo@latest" which pulls code from the npm registry and GitHub Actions like "uses: actions/checkout@v4"), so required external content is executed at runtime.