The Agent Skills Directory

Subprocess Command Execution: The skill includes utility scripts (scripts/run_eval.py, scripts/run_loop.py, scripts/improve_description.py) that utilize subprocess.run() and subprocess.Popen() to execute the claude CLI and other shell commands. While these are used within the context of benchmarking and testing the created skills, users should ensure they are running these in a secure environment.
Subagent Orchestration: The instructions guide the agent to spawn subagents for parallel execution of test cases and grading. This is a core feature of the skill's design to enable objective evaluation and human-in-the-loop iteration.
Local Web Server: The eval-viewer/generate_review.py script starts a temporary local HTTP server (default port 3117) to serve a self-contained HTML review page. This is intended to facilitate human review of test results in environments that support browser access.

skill-creator