vibe-3k is a professional workflow framework for Vibe Coding that optimizes the collaboration between human developers and AI coding agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install vibe-3k
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 vibe-3k 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 vibe-3k?
vibe-3k represents the third version of Vibe Coding Best Practices, designed to transform how developers interact with AI agents. It shifts the developer's role from a manual coder to a "head chef," where the human designs the menu and tastes the results while the AI handles the repetitive execution. By integrating these Openclaw Skills into your workflow, you ensure that AI-assisted development remains structured, secure, and highly productive.
The framework addresses the common pitfalls of AI development, such as context window overflow, repetitive repair loops, and security vulnerabilities. It emphasizes that while AI can handle the "vibe" of the code, humans must maintain control over architecture and safety-critical components.
vibe-3k Use Cases
- Launching new AI-assisted software projects with structured initializations.
- Coordinating complex multi-agent environments to prevent race conditions.
- Debugging and recovering from AI agents that have entered infinite fix-and-fail loops.
- Generating and maintaining standardized project rule files like CLAUDE.md or AGENTS.md.
- Reviewing and accepting large volumes of AI-generated code with a safety-first mindset.
How vibe-3k Works
- Perform a 5-minute pre-start checklist to define core features (P0/P1/P2) and technical constraints.
- Execute the PLAN phase using high-reasoning models to generate a comprehensive Design Doc.
- Initiate the ACT phase in a fresh session using lightweight models that follow the Design Doc for implementation.
- Maintain a LOG.md file to track progress, allowing the project state to be restored across different AI sessions.
- Conduct iterative Git commits for every functional unit to create a reliable safety net.
- Perform mandatory manual reviews for security-sensitive modules like authentication and database schemas.
vibe-3k Setup
To implement these Openclaw Skills in your development environment, organize your project structure as follows:
# Initialize project tracking files
touch LOG.md CLAUDE.md AGENTS.md
# Create a directory for task history and context preservation
mkdir -p memory/tasks/
# Set up the references directory for vibe-3k guidelines
mkdir -p references/
Ensure your AI agent is configured to read the .md files in the root directory to understand the project context and the vibe-3k operational standards.
vibe-3k Data Schema & Taxonomy
The vibe-3k framework organizes data and context using a specific taxonomy of Markdown files:
| Component | Description |
|---|---|
| Design Doc | The architectural blueprint generated during the PLAN phase. |
| LOG.md | A continuous log of progress and state to preserve context across session handoffs. |
| memory/tasks/ | A directory containing historical records of completed tasks and known issues. |
| Git Worktree | Used in multi-agent setups to manage parallel feature development without conflicts. |
vibe-3k Advanced Features
- PLAN/ACT Separation: Strict isolation of planning and execution phases to prevent context pollution.
- Multi-Agent Coordination: Specialized roles for PM, Dev, and QA agents with clear boundaries.
- 3-Strike Recovery Rule: Automatic strategy pivot if an AI agent fails to fix a bug in three attempts.
- Context Handover SOP: Detailed procedures for migrating state to a new session when context windows overflow.
- Automated Acceptance Checklists: Standardized verification steps for both automated and manual code reviews.
SKILL.md
Loading
Related Openclaw Skills
MiniMax MCP Call
A high-performance AI agent skill providing real-time web search and advanced image analysis via the MiniMax Coding Plan API.
Claw-News (Newsman)
An intelligent daily news briefing tool for Openclaw Skills that aggregates information via AI search, RSS feeds, and custom site trackers.
Class Seven Multi-Agent Development Team
Class Seven orchestrates a specialized multi-agent development team to handle complex software tasks from planning to deployment.
Catch My Skill
An automated utility to synchronize and track version differences for your local Openclaw Skills against GitHub and ClawHub repositories.
White Stone Memory
A structured, 5-category memory architecture for AI agents that optimizes context usage through on-demand loading and bilingual logging.
Code Detective
An interactive gamified experience that challenges developers to identify and fix bugs in code snippets across multiple difficulty levels.
Featured*
