The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check.py utilizes subprocess.run to execute Python-based validation and packaging tools.
Evidence: The _run(command) function in scripts/check.py executes commands constructed from sys.executable and local script paths like validate_skill.py, validate_evals.py, and package.py. These calls use list-based arguments without shell=True, which is a safe practice for developer tooling meant to audit and bundle skill assets.
[SAFE]: The skill demonstrates a strong security posture through its internal documentation and evaluation suite.
Evidence: The evals/evals.json file contains a specific test case (credential-network-boundary) that mandates the agent refuse any request to read ~/.env files or exfiltrate secrets to external services like Slack. This indicates the skill is designed to operate within strict safety boundaries regarding sensitive data access.

orchestrator