📜 Policy Editor

Visual interface for creating, editing, and testing safety policies with built-in templates.

Open Policy Editor

Ctrl+Shift+P → "Agent OS: Open Policy Editor"

Available Templates

🔒 Strict Security

Maximum security for production environments. Blocks external APIs, restricts file operations, enforces rate limits.

📋 SOC 2 Compliance

Enterprise compliance with audit logging, data encryption requirements, and access controls.

🇪🇺 GDPR Data Handling

EU data protection rules including PII detection, consent verification, and data minimization.

🛠️ Development

Permissive policy for local development. Allows most operations with warnings instead of blocks.

⏱️ Rate Limiting

API call restrictions to prevent abuse. Configurable limits per minute/hour.

Policy Structure

policy:
  name: "Policy Name"
  version: "1.0"
  rules:
    - name: "Rule name"
      condition: "when to apply"
      constraint: "what to check"
      action: "deny | warn | allow"
      message: "User-facing message"

Example: Custom Security Policy

policy:
  name: "Custom Security Policy"
  version: "1.0"
  rules:
    - name: "Block file writes outside /tmp"
      condition: "agent.action == 'file.write'"
      constraint: "not path.startsWith('/tmp/')"
      action: "deny"
      message: "File writes restricted to /tmp/"
    
    - name: "Block external APIs"
      condition: "agent.action == 'http.request'"
      constraint: "not url.host.endsWith('.internal.com')"
      action: "deny"
      message: "External API calls not allowed"
    
    - name: "Rate limit LLM calls"
      condition: "agent.action == 'llm.call'"
      constraint: "rate > 100 per minute"
      action: "throttle"
      message: "LLM calls rate limited"

Using the Editor

ActionHow To
Load TemplateSelect from dropdown at top
Edit PolicyModify YAML in editor
ValidateClick "Validate Policy" button
SaveClick "Save Policy" → choose location
ExportClick "Export" for different formats
💡 Tip: Use the Development template during local testing, then switch to Strict Security before deployment.