A comprehensive security hardening suite designed to protect AI agent workspaces from credential leaks and unauthorized access.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-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 agent-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 Agent Security?
Agent Security is a specialized toolkit designed to harden the environment where your AI agents operate. Because agents frequently interact with sensitive credentials, API keys, and private data, this skill provides the necessary guardrails to prevent accidental exposure. It integrates into the ecosystem of Openclaw Skills to provide automated auditing and monitoring, ensuring your local development environment remains secure.
By implementing this skill, developers can automate the detection of leaked secrets, verify that configuration files have the correct restrictive permissions, and monitor for unauthorized network patterns. It serves as a vital security layer for anyone deploying autonomous agents that require access to high-value infrastructure or private repositories.
Agent Security Use Cases
- Scanning workspace markdown and code files for hardcoded API keys or tokens.
- Auditing configuration file permissions to ensure sensitive data is not world-readable.
- Preventing the accidental inclusion of .env files or secret JSONs in Git version control.
- Tracking the age of credentials to enforce regular rotation policies within Openclaw Skills.
- Testing network connectivity to ensure the agent can only reach approved AI model endpoints.
How Agent Security Works
- The skill performs a recursive grep search across the workspace to identify patterns matching common API keys and private tokens.
- It executes a permission audit using system stat commands to verify that credential files are restricted to the current user (chmod 600).
- The system queries the Git index to check if any sensitive files are currently being tracked by version control.
- It calculates the delta between the current system time and the last modification date of security tokens to flag aged credentials.
- It initiates a network security check by performing silent curl requests to authorized hostnames to verify connection integrity.
Agent Security Setup
To deploy this security skill within your Openclaw Skills environment, ensure your workspace is configured and the necessary system utilities like grep and stat are available.
# Ensure your Openclaw workspace directory is accessible
mkdir -p ~/.openclaw/workspace/
# To perform an initial secret scan manually:
grep -rn --include='*.md' -E '(api[_-]?key|secret)' ~/.openclaw/workspace/
Agent Security Data Schema & Taxonomy
The Agent Security skill manages security metadata by auditing specific file paths and tracking permission states. It organizes its findings based on the following logic:
| Component | Target Path | Security Logic |
|---|---|---|
| Secret Scanner | ~/.openclaw/workspace/ | RegEx-based pattern matching for credentials |
| Permission Audit | ~/.config/ | Permission bit verification (target: 600) |
| Git Guard | Workspace root | Git ls-files filtering for sensitive keywords |
| Network Monitor | Outbound | HTTP response code validation for known API hosts |
Agent Security Advanced Features
- Heartbeat Integration: Automate daily security audits by adding the audit task to your HEARTBEAT.md file.
- Custom Secret Patterns: Extend the regex patterns to detect proprietary tokens or enterprise-specific keys within Openclaw Skills.
- Automated Permission Remediation: Scripts can be extended to automatically fix world-readable files.
- Network Whitelisting: Configurable host lists to restrict agent communication to verified endpoints only.
SKILL.md
Loading
Related Openclaw Skills
Agent Ping
Discover AI agents across the internet by scanning domains for Open Agent Discovery Protocol (OADP) signals.
Agent Network Scanner
A comprehensive scanning tool for locating AI agent networks, hubs, and active coordination platforms across various protocols.
Agent ID
Agent ID gives AI agents a persistent identity card and discovery signals to ensure they remain findable and reputable across platforms.
Agent Economy Starter Kit
A comprehensive toolkit to integrate AI agents into the ClawSwarm decentralized economy, enabling them to earn, negotiate, and build reputation.
Session Cost
A financial monitoring tool for Openclaw Skills that parses session transcripts to calculate per-model spend and enforce strict budget limits.
Agent Starter Kit
A comprehensive foundation that automates workspace setup, memory systems, and network discovery for new autonomous agents.
Featured*
