autoskill

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

• Potentially malicious external URL detected (high risk: 1.00). The skill sends runtime prompts to configured LLM endpoints (default local LM Studio at http://localhost:1234/v1, optional Anthropic at https://api.anthropic.com, or a user-supplied foundry endpoint) and parses the LLM responses (including "skill_body"/name) which are then written as SKILL.md drafts — i.e., remote HTTP responses directly control generated prompts/instructions during runtime.