The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Python subprocess module in scripts/platforms/claude_code.py, codex.py, and openclaw.py to execute agent CLI tools for trigger evaluation and benchmarking. This behavior is central to the skill's function and includes security measures like sandboxing where supported by the underlying platform.
[COMMAND_EXECUTION]: Additional subprocess calls in scripts/run_loop.py and scripts/improve_description.py facilitate the automated optimization of skill descriptions by calling optimizer runtimes.
[COMMAND_EXECUTION]: The eval-viewer/generate_review.py script uses shell commands to manage the lifecycle of the local evaluation viewer server and ensure ports are properly cleared.
[DATA_EXFILTRATION]: The eval-viewer component reads local workspace files and embeds them as Base64 data URIs in a local HTML file for review. This constitutes a data exposure surface for files within the project workspace, which is necessary for the qualitative review step of the development loop.

skill-creator