A runtime safety control plane that enforces deterministic security policies on shell commands, network requests, and file operations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install chainwatch
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 chainwatch 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 Chainwatch Safety Enforcement?
Chainwatch acts as a robust security layer for AI agents, providing a runtime control plane to intercept and evaluate commands before they are executed. By integrating this into your Openclaw Skills library, you can prevent agents from performing unauthorized or destructive actions. It ensures that every command—from simple file deletions to complex network requests—adheres to a predefined safety profile, providing developers with peace of mind when deploying autonomous agents.
The skill is specifically designed to handle high-risk operations by routing them through a policy engine. Whether the agent is attempting to modify system configurations or access sensitive credentials, Chainwatch provides a deterministic decision-making framework. This makes it an essential component for anyone building secure and reliable agents within the ecosystem of Openclaw Skills.
Chainwatch Safety Enforcement Use Cases
- Preventing accidental data destruction by intercepting commands like rm -rf or dd.
- Securing sensitive credentials by blocking unauthorized access to .env, .ssh, and .aws directories.
- Mitigating supply chain risks by monitoring package manager installations like npm or pip.
- Controlling network activity to prevent unauthorized outbound data transfers via curl or wget.
- Enforcing human-in-the-loop approval workflows for critical system or git configuration changes.
How Chainwatch Safety Enforcement Works
- The AI agent identifies a potentially sensitive or destructive command that requires policy enforcement.
- Instead of direct execution, the command is prefixed with the chainwatch exec utility using a specific security profile.
- The policy engine evaluates the command against established safety rules to determine if it should be allowed, blocked, or flagged for approval.
- If the command is safe, it executes normally; if blocked, Chainwatch returns a structured JSON object explaining the security violation.
- For commands requiring approval, the system pauses and provides an approval key for the user to authorize.
- Every evaluation and execution result is logged to an audit trail for forensic analysis and compliance monitoring.
Chainwatch Safety Enforcement Setup
To begin using this security feature within your Openclaw Skills environment, ensure the chainwatch binary is installed and accessible in your system path.
# Check if chainwatch is installed
chainwatch --version
# Perform a dry-run to test a specific command against the security policy
chainwatch evaluate --tool command --resource "rm -rf /tmp/data" --profile clawbot
Chainwatch Safety Enforcement Data Schema & Taxonomy
Chainwatch utilizes structured data formats to communicate safety decisions and maintain audit logs, ensuring compatibility with other Openclaw Skills.
| Component | Data Format | Description |
|---|---|---|
| Policy Response | JSON | Contains boolean blocked, string decision (allow/deny/require_approval), and reason. |
| Audit Logs | JSONL | Line-delimited JSON files tracking every command evaluation and execution result. |
| Profiles | Configuration | Named policy sets (like clawbot) that define the specific rules for different agent contexts. |
Chainwatch Safety Enforcement Advanced Features
- Dry-run evaluation mode to verify policy logic without performing any actual system changes.
- Granular approval workflow that allows users to authorize specific high-risk actions using unique approval keys.
- Comprehensive audit verification tools to inspect log files and ensure agent transparency.
- Profile-based enforcement which allows for different security tiers across various Openclaw Skills.
- Deep inspection of command patterns, preventing bypasses through clever shell syntax or aliasing.
SKILL.md
Loading
Related Openclaw Skills
Autonomous Execution
An AI agent skill designed to execute multi-step tasks autonomously while adhering to strict safety guardrails and confirmation protocols.
How To Skill
A meta-skill designed to teach developers and agents how to create, structure, and publish their own custom agentic capabilities.
Exoskeletons - AI Agent Identity Skill
Exoskeletons provide AI agents with a persistent onchain identity, reputation system, and secure marketplace built on the Base network.
Exponential CLI Skill
A high-performance CLI skill for the Exponential productivity platform, enabling seamless task creation, project tracking, and workspace management.
Host Hardening
A rapid security hardening skill to protect Linux servers running OpenClaw from unauthorized access and brute-force attacks.
Apple Health Data Analyzer
A privacy-focused CLI utility for analyzing Apple Health exports and extracting granular health metrics locally.
METADATA
Featured*
