The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses uv run to execute the nemo-evaluator-launcher CLI tool for managing the evaluation lifecycle, including running tasks, checking status, and listing past runs.
[REMOTE_CODE_EXECUTION]: The skill provides detailed instructions for using ssh and rsync to interact with remote clusters. These commands are necessary for monitoring live progress in logs and transferring evaluation artifacts for local analysis on the user's host.
[EXTERNAL_DOWNLOADS]: Instructions include installing the evaluator launcher directly from NVIDIA's official GitHub repository and fetching troubleshooting documentation via the GitHub API. These downloads are from the vendor's own infrastructure and are standard for this development tool.
[PROMPT_INJECTION]: The skill involves reading and analyzing various log files (client-*.log, server-*.log, slurm-*.log) generated during evaluations. This constitutes an indirect prompt injection surface as the agent processes data from external evaluation processes, which is the intended behavior for debugging and result analysis.

launching-evals