A specialized intent alias for VeriClaw designed to detect, diagnose, and fix AI hallucinations and false task completions through evidence-based correction loops.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install hallucination-correction-cn
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 hallucination-correction-cn 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 AI Hallucination Correction (CN)?
This specific entry in the Openclaw Skills registry serves as the primary Chinese language alias for VeriClaw. It is strategically designed to address the critical pain points of model overreach and false completion where an AI claims progress without actual results. By using these Openclaw Skills, developers and operators can transition from generic chat interactions to a rigorous, evidence-based verification framework.
The skill provides a dedicated pathway for hallucination remediation, ensuring that AI agents remain grounded in reality. It bridges the gap between raw model output and verifiable work by implementing a native correction companion that monitors for logical gaps, missing evidence, and fabricated facts.
AI Hallucination Correction (CN) Use Cases
- Identifying and correcting AI hallucination when a model fabricates facts or data points.
- Intervening during work hallucinations where the agent claims task completion without objective proof.
- Implementing evidence-based diagnosis for model overreaching or misinterpretation of context.
- Standardizing the remediation of model errors across different Chinese-language prompts and workflows.
How AI Hallucination Correction (CN) Works
- The system monitors for specific Chinese intent triggers related to hallucination and model correction.
- When a hallucination is detected, the skill initiates a verification loop to check for missing evidence.
- The operator performs a diagnosis using the provided diagnostic framework to identify the root cause of the error.
- An intervention is applied, which may involve re-running tasks, intercepting outputs, or upgrading processing logic.
- The results are verified against objective data to ensure the correction is successful and consistent.
AI Hallucination Correction (CN) Setup
To utilize this functionality within the Openclaw Skills ecosystem, you must install the canonical VeriClaw package. Run the following command in your terminal:
openclaw install vericlaw
After installation, you can access the hallucination correction features via the primary alias or the standard vericlaw interface.
AI Hallucination Correction (CN) Data Schema & Taxonomy
The skill organizes its correction data using a structured taxonomy to ensure reproducibility and learning. Below is the metadata structure used during the correction process:
| Data Component | Description |
|---|---|
| Evidence Log | Tracks files, outputs, or screenshots used to verify AI claims |
| Diagnosis Meta | Categorizes the type of hallucination (fact, task, or logic) |
| Intervention Action | Records the specific step taken to rectify the model error |
| Casebook Record | Stores the correction history for long-term model optimization |
AI Hallucination Correction (CN) Advanced Features
- Native Apple-native companion support for high-performance correction on macOS.
- Evidence-first verification loops that prioritize file-system and API truth over LLM assertions.
- Multi-path routing rules that direct discovery traffic to the most relevant correction surface.
- Custom review kits for accumulating real-world feedback and improving model reliability over time.
SKILL.md
Loading
Related Openclaw Skills
Umnix Jury System
A conceptual framework skill for transparent, multi-perspective AI dispute resolution featuring a 12-juror parallel AI panel and human oversight.
Umnix Constitution Engine Skill
A foundational governance framework defining mandates, permissions, and performance standards for autonomous AI agents.
HPM Solver
A mathematical optimization tool designed to calculate the most efficient purchase combinations in Harry Potter Magic Awakened.
Supply Chain Management Roadmap
A professional career development API that generates data-driven, personalized learning roadmaps for supply chain professionals.
Umnix AgentMemory Skill
A conceptual framework providing AI agents with persistent, structured memory across episodic, semantic, procedural, and relational types.
Google Sheets Maton Gateway
A powerful interface for AI agents to interact with Google Sheets via the Maton Gateway with full OAuth support.
Featured*
