MultiversX AI Skills (mx-ai-skills)

Empowering AI Agents with specialized knowledge, roles, and workflows for the MultiversX ecosystem.

This repository is a central hub for MultiversX AI Expertise. It provides a structured collection of specialized "Skills", "Global Workflows" (Roles), and "Documentation" designed to "equip" AI agents (like Antigravity) with the deep technical knowledge required to build, audit, and optimize on MultiversX.

---

🚀 Overview

AI Agents with specialized knowledge, roles, and workflows for the MultiversX ecosystem.

This repository is a central hub for MultiversX AI Expertise. It provides a structured collection of specialized "Skills", "Global Workflows" (Roles), and "Documentation" designed to equip AI agents with the deep technical knowledge required to build, audit, and optimize on MultiversX.

---

Quick Install

Install skills for any AI coding agent (Cursor, Windsurf, Codex, etc.):

npx openskills install multiversx/mx-ai-skills

Or using Vercel's skills CLI:

npx skills install multiversx/mx-ai-skills

---

Overview

Developing on MultiversX requires specific expertise in Rust (via multiversx-sc), WASM optimization, sharding awareness, and unique token standards (ESDT). This repository formalizes that expertise into machine-readable and agent-executable modules.

Core Value Props:

• Role-Based Execution: Instantiate specialized agents like MultiversX Smart Contract Developer or Full Stack Auditor.
• Granular Skills: Provide specific toolsets for tasks like Gas Optimization, Static Analysis, or Mandos Scenario Testing.
• Security First: Built-in rules for arithmetic safety, reentrancy protection, and MultiversX-specific "sharp edges".
• Documentation Engine: A curated library of protocol specs and best practices.

---

🛠 Repository Structure

The core logic resides in the antigravity/ directory, structured for easy ingestion by AI agents:

.
├── antigravity/
│   ├── global_workflows/    # Specialized Agent Roles (Roleplay instructions)
│   │   ├── rust-sc.md       # Expert Smart Contract Engineer
│   │   ├── mvx-auditor.md   # Full Stack Security Auditor
│   │   └── ...              # 20+ other specific roles
│   ├── skills/              # Targeted Technical Capabilities
│   │   ├── mvx_sc_best_practices/ # Optimization & Security guidelines
│   │   ├── mvx_sharp_edges/      # Known pitfalls and WASM behaviors
│   │   └── ...              # 19+ granular skills
│   ├── mvx_docs/            # Curated Protocol Documentation
│   └── GEMINI.md            # Global Workspace Rules & Core AI Personality
└── README.md                # You are here

---

🧩 Core Components

1. Global Workflows (Roles)

Roles defined in antigravity/global_workflows/ provide the "Persona" and "Protocol" for specific tasks.

• /mvx-smart-contract-developer: Focused on idiomatic Rust, gas efficiency, and storage mappers.
• /mvx-auditor: A rigorous framework for vulnerability research and system flow verification.
• /mvx-dapp-architect: Expert in React integration, sdk-dapp, and frontend security.

2. Specialized Skills

Skills in antigravity/skills/ are sets of instructions and checklists for specific activities:

• mvx_static_analysis: Patterns for finding unsafe code or floating-point arithmetic.
• consult_mvx_docs: High-speed lookup for MultiversX API and framework details.
• mvx_testing_handbook: Guidelines for Mandos scenarios and RustVM unit tests.

3. Global Rules (GEMINI.md)

The GEMINI.md file defines the fundamental behavioral constraints for AI agents in this workspace, ensuring they:

• Never hallucinate (Always ask for clarification).
• Prioritize arithmetic safety (Checked operations only).
• Follow the Checks-Effects-Interactions pattern strictly.

---

📖 How to Use

For AI Agents (like Antigravity)

These skills are automatically mapped to your workspace. You can invoke specific roles or skills using the @ symbol or by referencing the workflow name in your tasks.

For Developers

1. Read the Rules: Familiarize yourself with GEMINI.md to understand the standard of code expected.
2. Explore Skills: If you are performing a specific task (e.g., auditing), read the corresponding skill (e.g., antigravity/skills/mvx_dapp_audit/SKILL.md).
3. Invoke Roles: Use the slash commands defined in the workflows (e.g., /mvx-auditor) to guide the AI's behavior for complex tasks.

---

📜 Principles & Standards

• Code is Law: We prioritize security over convenience.
• Gas Efficiency: Every storage write is analyzed.
• Reproducibility: All builds and tests must be traceable and repeatable.
• Anti-Hallucination: If the documentation is missing, the agent MUST ask for data rather than guessing.

---

📄 License

• Universal Compatibility: Works with 30+ AI coding agents via OpenSkills standard
• Role-Based Execution: Instantiate specialized agents like MultiversX Smart Contract Developer or Full Stack Auditor
• Granular Skills: Specific toolsets for Security Auditing, Static Analysis, Property Testing, and more
• Security First: Built-in rules for arithmetic safety, reentrancy protection, and MultiversX-specific "sharp edges"

---

Repository Structure

mx-ai-skills/
├── skills/                     # Universal Skills (OpenSkills compatible)
│   ├── multiversx-smart-contracts/
│   ├── multiversx-dapp-frontend/
│   ├── multiversx-audit-context/
│   ├── multiversx-static-analysis/
│   └── ... (18 total skills)
│
├── antigravity/                # Antigravity-specific (legacy format)
│   ├── global_workflows/       # Agent Roles/Personas
│   ├── skills/                 # Original skills (underscore naming)
│   ├── mvx_docs/               # Curated documentation
│   └── GEMINI.md               # Global rules
│
├── README.md
└── LICENSE

---

Available Skills

Development

Skill	Description
multiversx-smart-contracts	Build, test, deploy smart contracts with Rust
multiversx-dapp-frontend	React dApp integration with sdk-dapp

Security & Auditing

Skill	Description
multiversx-audit-context	Build mental models before security audits
multiversx-entry-points	Map contract attack surface
multiversx-diff-review	Review changes between contract versions
multiversx-fix-verification	Verify vulnerability fixes
multiversx-dapp-audit	Audit frontend security
multiversx-static-analysis	Manual and automated code analysis
multiversx-constant-time	Verify timing-safe crypto operations
multiversx-variant-analysis	Find similar bugs after initial discovery

Testing & Quality

Skill	Description
multiversx-property-testing	Fuzzing and property-based testing
multiversx-spec-compliance	Verify implementations match specs
multiversx-project-culture	Assess codebase quality and maturity

Tooling & Reference

Skill	Description
multiversx-semgrep-creator	Create custom security scanning rules
multiversx-wasm-debug	Debug and optimize WASM binaries
multiversx-sharp-edges	Platform-specific gotchas catalog
multiversx-protocol-experts	Deep protocol knowledge (sharding, ESDT)
multiversx-clarification-expert	Ask targeted clarifying questions

---

Usage

For Any AI Coding Agent (Recommended)

# Install globally
npx openskills install multiversx/mx-ai-skills -g

# Or install in current project
npx openskills install multiversx/mx-ai-skills

For Antigravity

Use the original format in antigravity/skills/ with underscore naming:

antigravity/skills/mvx_static_analysis/SKILL.md

Manual Installation

# Clone to your project
git clone https://github.com/multiversx/mx-ai-skills.git .ai-skills

# Copy skills to your agent's expected location
cp -r .ai-skills/skills ~/.cursor/skills/        # Cursor
cp -r .ai-skills/skills ~/.agent/skills/         # Generic

---

Principles & Standards

• Never Hallucinate: Always ask for clarification when uncertain
• Security First: Checked arithmetic, input validation, CEI pattern
• Gas Efficiency: Minimize storage writes, optimize WASM size
• Reproducibility: All builds and tests must be traceable

---

Contributing

1. Add new skills to skills/<skill-name>/SKILL.md
2. Follow the YAML frontmatter format:

   ---
   name: skill-name
   description: Brief description. Use when [trigger conditions].
   ---

3. Use kebab-case for skill names
4. Include practical code examples and checklists

---

License

This repository is licensed under the MIT License - see the LICENSE file for details.