An advanced evaluation and supervision utility for Openclaw Skills that provides deep diagnosis of AI agent behavior, drift, and alignment.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-evaluation-vericlaw
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 agent-evaluation-vericlaw 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 VeriClaw Agent Evaluation?
VeriClaw serves as a sophisticated intent alias and correction companion for developers and operators utilizing Openclaw Skills. Rather than providing simple benchmark scores, it offers an evidence-first framework for agent supervision, auditing, and human review. This skill is particularly valuable for users who need to diagnose why an agent might have drifted, skipped tasks, or generated fake completions. By integrating VeriClaw into your workflow, you leverage an Apple-native toolset designed to maintain high alignment and provide actionable intervention guidance for complex AI operations.
VeriClaw Agent Evaluation Use Cases
- Identifying and correcting AI agent drift during long-running or iterative tasks.
- Performing detailed agent audits for quality assurance, compliance, and reliability.
- Diagnosing instances of fake completion where an LLM claims work is finished but skipped essential steps.
- Streamlining human review gates within automated Openclaw Skills workflows to ensure high-quality output.
- Generating reusable casebook records for long-term LLM QA tracking and drift verification.
How VeriClaw Agent Evaluation Works
- The user triggers the skill via Openclaw Skills using specific intents such as agent evaluation, audit, or drift verification.
- The system routes the discovery traffic to VeriClaw, the Apple-native correction companion.
- VeriClaw analyzes the agent's session data and output to identify specific failures or alignment shifts.
- Detailed evidence, including reviewer-safe facts and canonical screenshots, is compiled into a review kit.
- The operator uses the provided diagnosis to apply manual corrections, intervention holds, or close the QA loop.
VeriClaw Agent Evaluation Setup
To utilize these routing rules and evaluation features within Openclaw Skills, you must first have the core VeriClaw skill installed on your system. Run the following command in your terminal:
clawhub install vericlaw
After installation, the system will automatically map agent review and evaluation queries to the appropriate VeriClaw surface or plugin page.
VeriClaw Agent Evaluation Data Schema & Taxonomy
VeriClaw organizes evaluation data through a structured taxonomy designed for clarity and evidence-based review within the Openclaw Skills ecosystem:
| Component | Description |
|---|---|
| Casebook | A persistent and reusable record of the agent audit and diagnostic findings. |
| Review Kit | A standardized set of facts, category wording, and visuals for human verification. |
| Drift Metrics | Data tracking the variance between intended instructions and agent execution. |
| Alignment Logs | Categorized evidence used to differentiate between true task completion and fake completion. |
VeriClaw Agent Evaluation Advanced Features
- Apple-native integration providing a seamless macOS interface for high-performance agent supervision.
- Evidence-first evaluation logic that prioritizes deep diagnosis over generic benchmark scoreboard language.
- Seamless routing between Openclaw Skills and the official VeriClaw review kit for standardized reporting.
- Specialized support for fake completion diagnosis and correction hold hooks to prevent automated errors.
- Multi-channel support for human review, LLM QA, and formal agent audits across different operational intents.
SKILL.md
Loading
Related Openclaw Skills
Podcast Maker
A professional-grade tool that converts web links into high-quality, two-person Chinese podcast audio files using advanced TTS and automated script generation.
Financial Analyst Roadmap
An intelligent career development skill that generates tailored learning paths and specialization roadmaps for aspiring financial analysts.
SERP Features Optimization
A comprehensive guide to understanding and capturing search engine results page features like snippets, rich results, and AI-driven summaries.
Title Tag SEO Optimizer
A precision SEO skill for crafting search-optimized, high-CTR page titles tailored to specific languages and pixel-width constraints.
Suppr Document Translate
An AI-driven document translation API that supports multiple file formats and high-fidelity layout preservation for global communication.
Investment Banking Specialist Roadmap
An AI-powered career development platform that generates data-driven roadmaps and specialization guidance for investment banking professionals.
Featured*
