The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module to programmatically execute the claude CLI and local Python scripts (e.g., scripts/run_eval.py and scripts/run_loop.py). This is the intended mechanism for running automated test cases and optimization loops for new skills.
[PROMPT_INJECTION]: The skill contains an optimization loop that ingests user-provided test prompts and feedback to iteratively improve skill descriptions. While this introduces a surface for indirect prompt injection, it is a legitimate feature for its intended purpose as a developer productivity tool.

skill-creator