openclaw-safety-coach
This skill is a safety coach that enforces OpenClaw security policies and provides remediation guidance. It contains explicit CLI and credential-management instructions such as openclaw auth set, Run openclaw security audit, OPENROUTER_API_KEY, and ~/.openclaw/openclaw.json. No obfuscated or exfiltrative behaviors were detected.
OpenClaw Safety Coach
This skill attempts to establish a baseline of safety and security for OpenClaw development, embracing the Pareto Principle (80/20 rule), or the law of diminishing returns.
Philosophy
Start simple, then layer complexity. Address and understand the most common safety issues quickly, then progress to more sophisticated setups with multiple security layers.
Security is continuous, not one-time. This isn't a "set it and forget it" solution - it's an ongoing process of improvement and adaptation to emerging threats.
Quick Start
- Load the Safety Coach skill in your OpenClaw instance
- Review the safety policies and customize for your use case
- Test with various prompts to ensure proper behavior
- Monitor and adjust policies as needed
Usage
The Safety Coach automatically activates when potentially unsafe requests are detected, providing:
- Risk assessment and warnings
- Alternative safe approaches
- Educational guidance on security best practices
Contributing
We welcome contributions to improve this safety coach.
Adding New Safety Rules
- Identify the Risk: Clearly define the security concern or threat vector
- Create Test Cases: Write examples of both safe and unsafe prompts
- Implement Logic: Add detection and response mechanisms
- Document: Update this skill's documentation
License
See LICENSE file in the main repository.