A real-time security middleware that intercepts and validates every tool call made by your AI agent against custom regex-based policies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentgate-security
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentgate-security using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AgentGate - Enterprise Security Firewall?
AgentGate serves as a critical policy enforcement layer for Openclaw Skills, providing a robust firewall between your AI agent and sensitive system operations. By wrapping the agent's tool execution method, it ensures that every action—from shell commands to API calls—is evaluated against human-defined security rules before execution. This prevents common risks like command injection, data exfiltration, or accidental destructive actions caused by LLM hallucinations.
Built for enterprise reliability, this skill allows developers to maintain strict control over their agents without sacrificing performance. It integrates seamlessly into existing workflows using a lightweight SDK, connecting your agent to a centralized dashboard where you can manage API keys, audit logs, and approval queues in real-time.
AgentGate - Enterprise Security Firewall Use Cases
- Preventing destructive bash commands like rm -rf from being executed by an autonomous agent.
- Implementing spending limits and manual approvals for high-value financial transactions via Stripe.
- Restricting outbound HTTP requests to a specific whitelist of trusted domains for better security.
- Creating a human-in-the-loop workflow for sensitive operations using Telegram notifications.
How AgentGate - Enterprise Security Firewall Works
- The AgentGate middleware wraps the Openclaw agent's tool execution method during initialization.
- When the agent attempts a tool call, the SDK intercepts the tool name and serialized arguments.
- A secure POST request is sent to the AgentGate backend to evaluate the call against regex policies stored in Firestore.
- The system determines a decision: ALLOW, DENY, or REQUIRE_APPROVAL based on defined rule priorities.
- If approval is required, the execution pauses, and a webhook fires a notification to an operator via Telegram.
- Once a decision is logged, the agent either continues execution, receives a structured error, or waits for manual intervention.
AgentGate - Enterprise Security Firewall Setup
- Sign up for a free account at the AgentGate portal to obtain your credentials.
- Create a new agent entry in the dashboard and copy the generated API key.
- Install the guard package in your project directory:
npm install @agentgate/openclaw-guard
- Configure the SDK by wrapping your agent instance within your application code:
import OpenClaw from 'openclaw';
import { withAgentGate } from '@agentgate/openclaw-guard';
const agent = new OpenClaw({ model: 'claude-3-5-sonnet', tools: ['bash', 'fetch'] });
const securedAgent = withAgentGate(agent, {
apiKey: process.env.AGENTGATE_API_KEY,
endpoint: 'https://agent-gate-rho.vercel.app/api/evaluate-action'
});
AgentGate - Enterprise Security Firewall Data Schema & Taxonomy
AgentGate organizes its security policies and audit trails using the following metadata structure:
| Field | Type | Description |
|---|---|---|
| agentId | String | Unique identifier for the specific Openclaw agent |
| toolName | String | The name of the intercepted tool (e.g., bash, fetch, fs) |
| condition | Regex | The pattern used to match and validate tool arguments |
| ruleType | Enum | The outcome of a match: ALLOW, DENY, or REQUIRE_APPROVAL |
| priority | Integer | Evaluation order where lower numbers are processed first |
AgentGate - Enterprise Security Firewall Advanced Features
- AI Policy Wizard: Leverage Gemini to convert natural language descriptions into complex regex rules for your Openclaw Skills.
- Telegram Webhook Integration: Receive real-time approval requests on your mobile device for human-in-the-loop control.
- Real-time Audit Logging: Monitor every agent interaction through a live dashboard powered by Firestore onSnapshot.
- Sub-80ms Latency: High-performance evaluation ensures that security checks do not significantly impact agent response times.
SKILL.md
Loading
Related Openclaw Skills
Creative Ops Copilot
A production-ready planning agent that turns messy client briefs into structured VFX and motion design workflows.
Security Audit Enhanced
An advanced framework for automated security scanning, scoring, and hardening of AI agent environments.
Secrets Vault
A high-security, AES-256-GCM encrypted vault designed to manage API keys, database credentials, and sensitive tokens for developers.
Code Security Audit
A unified security audit toolkit for scanning vulnerabilities, detecting hardcoded secrets, and validating AI agent logic within modern codebases.
ClawHub CLI
A high-performance command-line interface for managing the lifecycle of Openclaw Skills, from discovery to deployment.
HWPX Document Processor (hwpx_my)
A specialized skill for generating and editing HWPX documents through ZIP-level template replacement and automation.
Featured*
