The Agent Skills Directory

[COMMAND_EXECUTION]: The package includes a testing suite (e.g., scripts/validate_skill_cli_tests.py) that uses the subprocess module to execute the skill's own validator script (validate_skill.py). This is a benign use of command execution intended for internal verification of the skill's structural logic.
[SAFE]: The skill implements strict structural validation (in scripts/check.py and scripts/reference_checks.py) to identify and reject the use of host-specific or global file paths (such as /Users or /home), ensuring the skill remains portable and secure.
[SAFE]: The instructions and evaluation cases explicitly define security boundaries, such as treating embedded instructions as untrusted data and refusing requests for secret collection or privilege escalation.

skill-creator