agents-docs

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires fetching and injecting live documentation into the agent context at runtime from https://developers.openai.com/codex/codex-manual.md and https://code.claude.com/docs/en/overview.md (and their domains), so those external URLs directly control prompts used to generate answers.