nemoclaw-user-overview

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly wires OpenShell-managed channel messaging (Telegram, Discord, Slack and similar platforms) into the sandboxed OpenClaw agent (see the "Channel messaging" sections in SKILL.md and references/how-it-works.md), so the agent ingests untrusted user-generated messages that could contain instructions affecting its behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly says the plugin "downloads the blueprint artifact, checks version compatibility, and verifies the digest" and then "resolves, verifies, and executes the blueprint as a subprocess" at runtime (e.g., artifacts referenced via https://github.com/NVIDIA/NemoClaw/releases), so remote code/artifacts are fetched and executed during runtime.