sdd-guardrails

What problem does it solve?

This Skill establishes robust security guardrails for AI agents, defining permission boundaries, implementing human-in-the-loop approval gates, ensuring audit trails, and enforcing workflow safety rules to prevent unauthorized actions and mitigate risks.

Core Features & Use Cases

• Define Permission Boundaries: Sets explicit rules for file system access, command execution, and API interactions based on the principle of least privilege.
• Implement Human-in-the-Loop Gates: Mandates human approval for critical, irreversible, or high-impact operations.
• Establish Audit Trails: Records all actions with detailed metadata for traceability and compliance.
• Enforce Workflow Safety: Integrates with security requirements (REQ-SEC-xxx) and SLO error budgets to control agent behavior dynamically.
• Use Case: Before an AI agent can deploy code to production, this Skill ensures a human approver reviews and authorizes the deployment, logs the action, and verifies it adheres to defined security policies.

Quick Start

Use the sdd-guardrails skill to generate security guardrails for the 'google-ad-report' specification.