The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.Popen within scripts/run_eval.py, scripts/improve_description.py, and eval-viewer/generate_review.py to interact with the claude CLI and system utilities like lsof. These calls are necessary for the skill's core functionality of executing evaluation runs and managing the local review server, and they follow safe implementation patterns by avoiding shell evaluation.
[DYNAMIC_EXECUTION]: The scripts/run_eval.py script dynamically generates temporary skill metadata files in the .claude/commands/ directory. This allows the tool to register and test new skill drafts during the evaluation process.
[INDIRECT_PROMPT_INJECTION]: As a testing tool, the skill is designed to ingest and process arbitrary user-provided prompts. While this is an attack surface, it is fundamental to the skill's purpose as an evaluation framework, and the prompts are passed to subagents using secure CLI argument patterns.
[EXTERNAL_DOWNLOADS]: The eval-viewer/viewer.html file includes a reference to the SheetJS library hosted on a public CDN (cdn.sheetjs.com). This well-known library is used to render spreadsheet outputs directly in the review interface for the user.

skill-creator