bleu

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly instructs re-fetching Claude Code documentation at runtime (e.g., https://code.claude.com/docs/en/hooks and related https://code.claude.com/docs/en/* links) before generating settings/agent files and hooks, meaning external content would be fetched during execution and used to build agent prompts/configs that control behavior—so this is a runtime external dependency that directly shapes agent instructions.