auth-md

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime “validate” mode can fetch live discovery documents from outsider-controlled URLs (e.g., user-supplied base_url/auth.md URL) and ingest their JSON into the agent’s LLM context, which is an indirect prompt-injection risk from public web content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs agents to fetch a service's auth.md at runtime (for example: https://acmenotes.com/auth.md, and more generally https:///auth.md as referenced in the Authorization Server metadata) and to use that fetched document (and the well-known discovery docs) as authoritative runtime instructions for registration, so this is a runtime external dependency that directly controls agent prompts/behavior.