tao-run-on-slurm

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime reads SLURM job log files over SSH from <job_dir>/slurm-logs/<slurm_job_name>-<slurm_job_id>/main.out and main.err, where the log text is produced by the user’s job/container (i.e., outsider-authored free text relative to the agent’s control), and that text is then ingested into the agent context for status/retry decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's preflight/install steps perform runtime pip installs (e.g., the suggested REPO git+https://gitlab-master.nvidia.com/nvidia-tao-toolkit/tao-sdk.git and the python -m pip install "$PIN" path in references/slurm-ssh-credentials.md), which fetch and execute remote code during skill runtime, so the git+https://gitlab-master.nvidia.com/nvidia-tao-toolkit/tao-sdk.git URL is a high-confidence runtime external dependency.