VibeCoding Pro is a high-performance AI coding framework that replaces biased self-evaluation with an independent, browser-based QA agent to ensure code actually works.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install vibecoding-pro
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 vibecoding-pro 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 VibeCoding Pro?
VibeCoding Pro transforms the standard AI coding process from a write-and-hope approach into a disciplined, iterative engineering workflow. Inspired by Anthropic's 2026 engineering research and Generative Adversarial Network (GAN) theory, it structurally eliminates AI self-evaluation bias. By decoupling the creative Generator agent from an independent Evaluator agent, it ensures that code is validated through real-world interactions rather than just internal logic.
The core strength of this skill lies in its architectural separation. The Evaluator agent never sees the Generator's code or reasoning; it only interacts with the live, deployed artifact using browser automation. This adversarial tension forces the AI to move beyond vibe coding and produce software that meets rigorous functional and interaction-based acceptance criteria.
VibeCoding Pro Use Cases
- Building React or H5 Web UIs that require real interaction testing via Playwright.
- Developing multi-step form flows like wizards, checkouts, and onboarding sequences.
- Validating API and frontend integrations by checking real-time database states.
- Managing multi-round AI-assisted development sprints for production-critical features.
- Reducing AI slop by enforcing a structured scoring rubric and feedback loop.
How VibeCoding Pro Works
- The process begins with a Spec Contract, which serves as the single source of truth for the Evaluator.
- The Generator Agent receives the spec and previous feedback to build and deploy an artifact.
- The Evaluator Agent accesses the deployed URL using a real browser, completely independent of the source code.
- Using Playwright or similar tools, the Evaluator executes test scenarios, clicks through the UI, and captures evidence.
- The Evaluator generates a structured JSON feedback report with a score (0-100) based on a weighted rubric.
- If the score meets the predefined threshold, the loop finishes; otherwise, it triggers a new iteration for the Generator.
VibeCoding Pro Setup
To get started with VibeCoding Pro, you can implement the iteration loop using the provided Python scripts. Ensure you have your environment configured for Openclaw Skills and the Playwright MCP tool.
# Install dependencies
pip install playwright
playwright install
# Run the iteration loop with a specific threshold
python iteration_loop.py --spec spec.md --url http://localhost:3000 --threshold 85 --rounds 15
VibeCoding Pro Data Schema & Taxonomy
The skill organizes its workflow around a strict metadata and document taxonomy to maintain agent independence.
| File/Artifact | Description | Owner |
|---|---|---|
spec.md |
The Functional and Interaction requirements contract. | User |
Generator Output |
The source code and deployed artifact files. | Generator |
evaluator_report.json |
Structured scoring, screenshots, and failure logs. | Evaluator |
iteration_history |
The log of feedback used by the Generator to improve. | System |
VibeCoding Pro Advanced Features
- Support for four architecture variants: Sequential, Parallel, Staged, and Human-in-the-loop.
- Plateau detection and strategy switching to prevent the Generator from getting stuck in local minima.
- Automated evaluator calibration to prevent score drift and ensure consistent quality standards.
- Integration with Openclaw Skills subagents using the Playwright MCP tool for real-world UI interaction.
- Custom scoring rubrics that weight functional completeness, edge case handling, and craft quality.
SKILL.md
Loading
Related Openclaw Skills
MOVA Intent Calibration
A structured, 9-step guided thinking process to transform raw user requests into explicit, testable, and responsibility-bearing intents.
WSL Screenshot
A specialized bridge that allows WSL2 users to capture Windows desktop screenshots directly from the Linux terminal via PowerShell invocation.
Skill Trend Analysis
A professional analytics skill designed to track growth trends, rankings, and performance metrics for Openclaw Skills through structured API queries.
Another Android Device Control
A professional-grade MCP server skill that enables AI agents to fully control, automate, and interact with Android devices via adb and scrcpy.
TQSDK Read: Futures Market Data Connector
A professional financial data skill for accessing real-time futures quotes, options pricing, and historical K-line data via the TQSDK framework.
EdgeBets Sports Simulation
EdgeBets is a high-performance sports simulation engine that uses Monte Carlo analysis to identify betting edges and predict game outcomes.
Featured*
