The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/benchmark_command.py utility uses subprocess.run to execute commands for performance measurement. This is a primary function of the skill. The instructions strictly limit its use to workloads explicitly approved by the user during the mandatory checkpoint process.
[SAFE_PRACTICES]: The skill instructions proactively address security and integrity by requiring the agent to establish a 'contract' of preserved behavior and explicitly forbidding the removal of security or validation logic to achieve performance gains.
[DATA_EXPOSURE]: Analysis of the provided scripts confirms they do not contain hardcoded credentials, API keys, or logic for external data transmission. All operations are performed locally on the user's project files and benchmark results.

speedup-proof