evaluations

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly calls fetch_langwatch_docs at runtime to load external documentation (e.g., https://langwatch.ai/docs/evaluations/overview.md), and that fetched content is used to drive prompt/evaluator/guardrail setup—meaning remote content can directly control agent instructions at runtime.