The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references configuration and guidelines from trusted organizations including Anthropic's official skills repository and specialized tools like Claude Code.
[COMMAND_EXECUTION]: Several scripts (scripts/run_eval.py, scripts/test_smoke.py, eval-viewer/generate_review.py) use subprocess calls to execute system utilities (e.g., uv, lsof, claude). These operations are intrinsic to the skill's purpose of automating skill development and execution testing.
[REMOTE_CODE_EXECUTION]: While the skill instructions mention installing the uv package manager via a curl pipe to bash (curl -LsSf https://astral.sh/uv/install.sh | sh), this is documented as a standard prerequisite for the toolchain and targets a known official installation source.
[DATA_EXPOSURE]: The skill includes automated checks to prevent developers from accidentally including secrets or absolute system paths in their own skill files, promoting security best practices during the development lifecycle.

skill-conductor