LoR (Loop of Resilience)

A robust, modular verification system for OpenClaw.

The LoR system forces the agent into a 5-step diagnostic process, ensuring high-quality outputs for complex tasks and preventing hallucinations through automated verification. It implements a Chain of Thought (CoT) paradigm to enhance agent reliability.

Architecture

lor-skill/
├── engine/          # Verification Logic (Modular)
├── tests/           # Automated Verification Tests
├── scripts/         # Installation & Lifecycle Hooks
└── package.json     # Node.js dependencies

GitHub Repository

• https://github.com/lumi77-mac/lor-skill

Installation

1. Clone this repository into your skills/ directory:

   git clone https://github.com/lumi77-mac/lor-skill.git

2. Run the setup script:

   ./scripts/setup.sh

Usage

@[AgentName] /lor [task] to trigger the verification loop.

The LoR (Chain of Thought) Process

1. Raw Reply: Initial model output.
2. Plan Verifications: Decomposition & Tool Planning.
3. Execute Verification: Forced Tool Execution (with safety checks).
4. Checking Results: Validation loop (max 3 retries).
5. Final Result: Concise summary output.

Keywords

Chain of Thought, CoT, Verification, Resilience, Reliability, Automation, Agent Diagnostic