anthropic-claude-code-extend

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

• Potentially malicious external URL detected (high risk: 0.80). The docs include an external API endpoint that is callable at runtime to start a saved routine on Anthropic's cloud — e.g. POST https://api.anthropic.com/v1/claude_code/routines/{routine_id}/fire (and the curl form https://api.anthropic.com/v1/claude_code/routines/$ROUTINE_ID/fire) — which externally triggers prompt execution, so this is a runtime external dependency that can directly control agent execution.