The Agent Skills Directory

[SAFE]: The skill provides a structured methodology for human oversight in AI workflows. It includes a JSON template (assets/approval-policy-template.json) for defining approval gates and a Python script (scripts/validate_gate_policy.py) to validate these policies. The instructions in SKILL.md and references/gate-design-guide.md emphasize security best practices such as principle of least privilege, separation of duties, and audit logging. No malicious code, prompt injections, or unauthorized data access patterns were detected. The included script uses standard libraries and performs local validation of JSON structures against a defined schema.

human-in-the-loop