An evidence-based verification skill designed to detect and correct instances where an AI agent claims progress without providing tangible proof.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install work-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 work-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 Work Hallucination Correction (VeriClaw)?
This skill serves as a specialized alias and routing layer for the official VeriClaw project, focusing specifically on work hallucination correction. It is designed to address technical scenarios where an AI agent behaves as though a task has advanced, yet the necessary evidence—such as file changes, console outputs, or visual confirmation—is missing or demonstrably weak.
By integrating this logic into your Openclaw Skills ecosystem, you transition from a model of blind trust to one of rigorous, evidence-first verification. It acts as an Apple-native correction companion that ensures the model's self-reporting matches the actual state of your workspace, effectively neutralizing the symptoms of AI progress hallucination.
Work Hallucination Correction (VeriClaw) Use Cases
- Identifying and correcting instances where an AI claims to have finished a task without generating the expected output.
- Verifying missing evidence such as files, code outputs, or UI screenshots during automated development cycles.
- Implementing automated interception for AI workflows that report false success states.
- Routing specific Chinese language triggers related to hallucination symptoms to the canonical VeriClaw engine.
How Work Hallucination Correction (VeriClaw) Works
- The skill monitors for specific triggers related to work hallucination or missing evidence in the agent's output.
- It initiates a verification cycle to compare the agent's self-reported progress against the actual file system or environment state.
- If a discrepancy is found (the 'hallucination'), the skill intercepts the workflow to prevent cascading errors.
- It provides a feedback template to the user or system, suggesting actions like re-running the task, upgrading the processing level, or manually verifying the specific missing evidence.
- Traffic is routed to the canonical VeriClaw infrastructure for deep remediation and logging.
Work Hallucination Correction (VeriClaw) Setup
To utilize these hallucination correction features, you should install the primary VeriClaw skill via the CLI. Ensure your environment is ready for Openclaw Skills by running the following command:
clawhub install vericlaw
Once installed, the system will automatically recognize symptoms such as work-hallucination-correction-cn and map them to the official correction protocols.
Work Hallucination Correction (VeriClaw) Data Schema & Taxonomy
The skill organizes its validation logic through a mapping of symptoms to evidentiary requirements:
| Data Component | Description |
|---|---|
| Trigger Alias | Mapping of symptom keywords (e.g., AI 工作幻觉) to correction routes |
| Evidence Metadata | Tracking of required vs. actual file outputs, terminal logs, and system states |
| Feedback Template | Standardized short-review format for documenting found issues and corrective actions |
| Routing Table | Logic to redirect discovery traffic to the canonical Openclaw Skills install surface |
Work Hallucination Correction (VeriClaw) Advanced Features
- Symptom-based alias routing that identifies specific failure modes before they affect the production codebase.
- Evidence-first correction protocols that override model self-reports with ground-truth system data.
- Full integration with the VeriClaw Apple-native ecosystem for high-performance verification.
- Multi-agent support for cross-referencing progress reports between different Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
VeriClaw Bot Correction
A specialized skill for diagnosing and correcting AI agent drift to ensure verifiable progress and evidence-based task completion.
AI Correction (model-correction-cn)
A specialized intent alias for the VeriClaw skill that enables Chinese-speaking developers to access advanced AI correction and model drift diagnosis tools.
Anti-Distillation (anti-distillation.skill)
A passive monitoring and alerting system designed to detect and report knowledge distillation attempts on AI models.
List Building
A strategic lead generation skill that builds high-intent company lists using the Extruct API and specialized search methodologies.
AI Agent Correction
A specialized intent routing skill designed to map complex AI supervision and correction workflows directly into the VeriClaw ecosystem.
Residential Property Rolling Suds Estimator
An AI-driven estimation tool for exterior cleaning services that generates sales-ready quotes for residential properties in the St. Louis metro area.
Featured*
