openclaw-egress

What This Skill Does

OpenClaw Egress is a critical security utility designed to provide Data Loss Prevention (DLP) for AI agent workspaces. As AI agents gain the ability to interact with files and local environments, the risk of sensitive data exfiltration increases significantly. This skill mitigates that risk by scanning your workspace for unauthorized outbound network activity, identifying potentially malicious API endpoints, and flagging suspicious code patterns that could leak your credentials, conversation history, or proprietary codebase.

Unlike external security tools, OpenClaw Egress runs entirely locally using only the Python standard library. It does not phone home, nor does it require any third-party dependencies, ensuring your security analysis does not introduce new attack vectors. It provides a multi-tiered threat detection system, categorizing findings into CRITICAL, HIGH, WARNING, and INFO tiers based on the nature of the network call and the destination.

Installation

To integrate OpenClaw Egress into your workflow, use the standard ClawHub installer. Open your terminal or your agent's command interface and run:

clawhub install openclaw/skills/skills/atlaspa/openclaw-egress

Once installed, the tool is ready for immediate use. Because it operates on local files, ensure your agent has appropriate read permissions for the directories you wish to scan.

Use Cases

• Pre-Commit Security Audit: Scan your agent's workspace before committing code to ensure no hardcoded secrets have been added that might be leaked via network requests.
• Supply Chain Protection: When using third-party agent skills, use this tool to ensure those skills are not making unauthorized calls to dynamic DNS services or request catchers.
• Security Compliance: Regularly audit your workspace against external endpoints to identify and catalog all external dependencies being called by your scripts.
• Incident Response: If you suspect an agent has been compromised, use the status and scan commands to immediately identify the culprit script making malicious network calls.

Example Prompts

1. "@openclaw-egress, perform a full scan of the current directory to see if any scripts are sending data to external servers."
2. "@openclaw-egress, list every external domain referenced in the current codebase to help me audit my project's dependencies."
3. "@openclaw-egress, check the current workspace for high-risk network calls like hardcoded webhooks or potential exfiltration endpoints."

Tips & Limitations

• Execution Context: Since this skill uses Python's ast or file-parsing mechanisms, it is best at detecting static code patterns. It may not catch heavily obfuscated or dynamically constructed URLs generated at runtime.
• False Positives: If you have legitimate integrations (like legitimate API wrappers), you may see INFO or WARNING flags. Use the tool to map these and ensure your environment is configured to your organization's allowlist.
• Maintenance: Run the scan frequently, especially after installing new agent skills or updating your workspace configuration.