The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local Python modules and shell commands (e.g., python -m scripts.aggregate_benchmark, nohup, kill, cp) to manage the lifecycle of skill creation, run subagent evaluations, and launch a local review server. These operations are restricted to the local environment and the skill's own workspace.
[DYNAMIC_EXECUTION]: The instructions encourage users to create helper scripts within a scripts/ directory for repetitive tasks and to write programmatic assertion scripts for automated grading. This dynamic script generation is intended for developer automation and testing.
[DATA_EXPOSURE]: The skill manages data within a workspace directory, including test prompts, execution results, and user feedback. It generates temporary HTML files for local review, which is a standard pattern for providing qualitative feedback in development environments.

skill-creator