exec-slurm-compile

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches container images from external registries via the enroot-import script (docker://<image_url>) and runs those containers, and its required workflow (Step 0 + scripts/compile.slurm and Step 4) actively reads and interprets build logs from the job to drive monitoring and next actions, so untrusted third‑party content can materially influence the agent.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill’s enroot-import step fetches remote container images at runtime (e.g., docker://urm.nvidia.com/sw-tensorrt-docker/tensorrt-llm:...) which are required to produce the .sqsh image and will supply/execute remote code inside the container, so this is a runtime external dependency that can execute code.