nemo-rl-e2e-testing

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

• Third-party content exposure detected (high risk: 0.90). This SKILL.md explicitly instructs cloning a public GitHub repo and loading public Hugging Face model/tokenizer snapshots (e.g., policy.model_name=Qwen/... and AutoConfig.from_pretrained with trust_remote_code=True), so it fetches and executes/interprets untrusted, user-provided third-party content that directly affects test runs and pass/fail decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs cloning and importing the upstream NeMo-RL repository (https://github.com/NVIDIA-NeMo/RL.git), which fetches remote code that will be executed at runtime and is relied upon as a required dependency.