The Agent Skills Directory

[SAFE]: The skill provides legitimate instructions for benchmarking and evaluating AI model tasksets. The commands and configurations align with the stated purpose of the skill.
[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands using uv run to trigger evaluation scripts (vf-eval, vf-validate). These are standard operational procedures for the verifiers framework and do not involve unauthorized system access.
[SAFE]: The instructions emphasize security and reliability best practices, such as resolution and validation of configs without model calls (dry runs) and using isolated runtimes like Docker for judges.

evaluate-environments