OpenClaw Guardrails API

Use these endpoints to vet OpenClaw skills, authorize runtime actions, and manage project-level guardrail policy.

Base URL

https://api.vettly.dev

Authentication

All requests require an auth bearer token (API key or session token).

Authorization: Bearer vettly_xxxxxxxxxxxxx

Endpoints

Vet a skill

POST /v1/openclaw/guardrails/skill-vetting

Request

{
  "skillId": "net-utils-v1",
  "name": "Network Utilities",
  "source": "clawhub",
  "prompt": "Fetch remote diagnostics and summarize output.",
  "requestedPermissions": ["network_access", "run_shell"],
  "hasChecksum": true,
  "signedPublisher": true,
  "maintainerReputation": "trusted"
}

Response

{
  "decisionId": "2c1f93d7-5370-48f1-bf1e-3748d9e4061a",
  "action": "warn",
  "riskScore": 38,
  "reasons": [
    "Skill comes from a public marketplace source.",
    "Skill requests high-risk permission: run_shell."
  ],
  "categories": ["marketplace_source", "high_privilege_permission"],
  "requiresApproval": true
}

Authorize a runtime action

POST /v1/openclaw/guardrails/action-authorize

Request

{
  "actionType": "exec",
  "command": "sudo rm -rf /tmp/old-cache",
  "targetPath": "/tmp/old-cache",
  "requestedPermissions": ["run_shell", "delete_file"]
}

If policy is omitted, Vettly applies project-level OpenClaw policy (if set), then fallback defaults.

Response

{
  "decisionId": "db1ac66f-2036-42d7-b728-9ec8a3e258f5",
  "action": "flag",
  "riskScore": 57,
  "reasons": [
    "Action type exec is sensitive by default.",
    "Command includes high-risk shell or exfiltration behavior."
  ],
  "categories": ["sensitive_action", "dangerous_command"],
  "requiresApproval": true
}

Get project policy

GET /v1/openclaw/guardrails/policy?projectId={projectId}

Response

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "policy": {
    "sensitivePathPrefixes": ["/etc", "/usr", ".ssh", ".env"],
    "networkAllowlist": ["api.openai.com", "api.anthropic.com"],
    "blockShellPatterns": ["rm -rf", "curl | sh", "wget | sh"],
    "requireApprovalScore": 40
  }
}

Get policy history

GET /v1/openclaw/guardrails/policy/history?projectId={projectId}

Response

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "history": [
    {
      "version": "2f64d4bb7d57",
      "updatedAt": "2026-02-12T20:19:00.201Z",
      "updatedBy": "user_123",
      "policy": {
        "sensitivePathPrefixes": ["/etc", "/usr", ".ssh", ".env"],
        "networkAllowlist": ["api.openai.com"],
        "blockShellPatterns": ["rm -rf", "curl | sh"],
        "requireApprovalScore": 45
      }
    }
  ]
}

Roll back policy version

POST /v1/openclaw/guardrails/policy/rollback

Request

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "version": "2f64d4bb7d57"
}

Get project metrics

GET /v1/openclaw/guardrails/metrics?projectId={projectId}&days=30

Response

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "days": 30,
  "total": 128,
  "blocked": 9,
  "approvalRequired": 22,
  "blockRate": 0.0703,
  "approvalRate": 0.1719,
  "skillVetting": 41,
  "runtimeAuthorization": 87,
  "daily": [
    { "date": "2026-02-10", "total": 12, "blocked": 1, "approvalRequired": 3 }
  ]
}

Update project policy

PUT /v1/openclaw/guardrails/policy

Request

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "policy": {
    "sensitivePathPrefixes": ["/etc", "/usr", "/home"],
    "networkAllowlist": ["api.openai.com"],
    "blockShellPatterns": ["rm -rf", "curl | sh"],
    "requireApprovalScore": 45
  }
}

Response

{
  "projectId": "6a6f7cdd-6f2d-4fb3-a3b4-9156607906fe",
  "policy": {
    "sensitivePathPrefixes": ["/etc", "/usr", "/home"],
    "networkAllowlist": ["api.openai.com"],
    "blockShellPatterns": ["rm -rf", "curl | sh"],
    "requireApprovalScore": 45
  }
}

Webhook events

These guardrail endpoints emit:

• openclaw.guardrail.blocked
• openclaw.guardrail.approval_required

See /api/webhooks for payload examples.