nemoclaw-user-configure-inference

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

• Third-party content exposure detected (high risk: 1.00). The onboarding wizard explicitly probes and sends real inference requests to user-supplied OpenAI-/Anthropic-compatible endpoints (see "Other OpenAI-compatible endpoint" and validation steps that test /responses vs /chat/completions in SKILL.md), and those untrusted third‑party model responses are parsed to decide API paths and enable tool-calling behavior, which can materially influence the agent's runtime actions.