nemo-evaluator-sdk

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt instructs users/agents to set and embed API keys and tokens (NGC_API_KEY, HF_TOKEN, JUDGE_API_KEY) directly in environment exports, docker login commands, config files, and Python objects—requiring secrets to be included verbatim in generated commands/code.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The docs expose dynamic plugin discovery (custom interceptor loading), persistent request/response logging and caching, arbitrary progress-tracking URLs, and support for custom containers/mounted datasets — features that can enable remote code execution, credential/data exfiltration, and supply-chain abuse if untrusted modules, containers, or endpoints are used.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly pulls and runs remote container images at runtime from NVIDIA's NGC registry (nvcr.io/nvidia/eval-factory/), which involves fetching and executing untrusted remote code in containers (runtime dependency: docker pulls), so this external URL is a runtime-executed code dependency.