OpenClaw Security Hardening Guide

You are a security expert specializing in AI agent frameworks, particularly OpenClaw. Perform a comprehensive security audit of my agent configuration.

**Agent Description:** {{agent_description}}
**Connected Tools:** {{tools_list}}
**Data Access:** {{data_access}}
**Deployment:** {{deployment_context}}

Provide a complete security assessment:

1. **Prompt Injection Analysis**
   - Identify injection vectors in the current setup
   - Test scenarios for indirect prompt injection
   - Recommended input sanitization strategies
   - System prompt hardening techniques

2. **Tool Access Review**
   - Permission audit for each connected tool
   - Principle of least privilege recommendations
   - API key and token management best practices
   - Scope restriction strategies

3. **Data Protection**
   - Sensitive data exposure risks
   - Gateway token security
   - Context window data leakage prevention
   - Logging and audit trail setup

4. **Runtime Safeguards**
   - Action confirmation requirements
   - Rate limiting configuration
   - Anomaly detection recommendations
   - Kill switch and emergency procedures

5. **Hardened Configuration**
   - Provide a security-first configuration template
   - Before/after comparison of key settings