The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.Popen in several scripts (run_eval.py, improve_description.py, generate_review.py) to invoke the claude CLI and system utilities like lsof. This is necessary for its primary purpose of running evaluation loops and managing local servers for the benchmark viewer.
[DYNAMIC_EXECUTION]: The skill generates and saves eval_set.json and various grading.json files based on agent-generated content. It also dynamically builds and writes HTML files to the local file system to provide a review interface for users. These operations are intended for the skill development lifecycle.
[METADATA_POISONING]: The SKILL.md frontmatter includes standard fields (name, description). The description field is used by the skill to explain its purpose and is itself a target for optimization by the skill's own scripts, but no malicious intent was found in the provided instructions.

skill-creator