Battle-tested AI agent skills for Solidity smart contract development
Equip your coding agent with professional-grade Solidity practices — from coding standards to DeFi security.
EVM Compatible
solidity-agent-kit is a collection of AI agent skills that supercharge your coding agent with professional-grade Solidity development knowledge. Instead of writing one-off prompts or hoping your AI agent knows best practices, you install these skills once and get consistent, battle-tested behaviors across every project.
Equip your AI coding agent with battle-tested Solidity development practices — from coding standards to DeFi security.
Without solidity-agent-kit With solidity-agent-kit
───────────────────────── ─────────────────────────
AI agent writes code... AI agent invokes /solidity-coding
❌ inconsistent style ✅ enforces NatSpec, custom errors
❌ misses security checks ✅ runs /solidity-security rules
❌ no deployment verification ✅ pre-deploy checklist enforced
❌ weak test coverage ✅ fuzz tests, 100% branch coverage
❌ DeFi vulnerabilities ✅ anti-flash-loan, anti-whale guards
- Install
- Skills Included
- EVM Chain Support
- Architecture
- Supported Agents
- Tech Stack
- Project Structure
- License
# Interactive — choose which skills to install
npx skills add 0xlayerghost/solidity-agent-kit
# Install all skills at once (no prompts)
npx skills add 0xlayerghost/solidity-agent-kit -yAuto-detects all AI agents on your machine and installs for each one. Creates .agents/, .claude/, .cursor/, .windsurf/ folders — .agents/ holds the actual files, others are symlinks.
# Claude Code
npx skills add 0xlayerghost/solidity-agent-kit -y --agent claude-code
# Cursor
npx skills add 0xlayerghost/solidity-agent-kit -y --agent cursor
# Windsurf
npx skills add 0xlayerghost/solidity-agent-kit -y --agent windsurfCopy the CLAUDE.md template to your project root so skills are auto-invoked — no /slash-commands needed:
curl -sL https://raw.githubusercontent.com/0xlayerghost/solidity-agent-kit/main/CLAUDE.md.template -o CLAUDE.mdOr manually copy CLAUDE.md.template → CLAUDE.md in your project root.
CLAUDE.mdis loaded at the start of every Claude Code session. It tells your agent when to invoke each skill automatically.
Add to .gitignore to keep your repo clean:
# Agent skills (installed locally, not committed)
.agents/
.claude/
.cursor/
.trae/
.windsurf/npx skills remove 0xlayerghost/solidity-agent-kit| # | Skill | Trigger | Description | Link |
|---|---|---|---|---|
| 1 | solidity-coding | Writing / editing .sol files |
Pragma, naming, NatSpec, custom errors, anti-patterns, OZ imports | View |
| 2 | solidity-security | Writing / editing any contract | Private key protection, access control, reentrancy guards, gas limits | View |
| 3 | solidity-testing | Writing / editing *.t.sol |
Test layout, coverage ≥90%, fuzz testing, invariant tests with Foundry | View |
| 4 | solidity-deploy | Writing *.s.sol / deploying |
Pre-deploy checklist, env vars, constructor args, post-deploy workflow | View |
| 5 | solidity-checklist | Before any on-chain op (cast send, --broadcast) |
6-layer verification: permissions, dependencies, params, security, testing, knowledge capture | View |
| 6 | solidity-debug | Debugging failed txs | Cast calldata decoding, revert analysis, gas diagnosis, trace reading | View |
| 7 | solidity-audit | Security reviews | Manual audit checklist: storage, access control, arithmetic, ERC compatibility | View |
| 8 | defi-security | DeFi protocol work | Anti-whale, anti-flash-loan, launch checklist, emergency pause, MEV defense | View |
| 9 | git-workflow | Before commits / PRs | Conventional commits, branch naming, PR templates for Solidity projects | View |
| 10 | claude-code-usage | After every /clear |
Context recovery, subagent strategy, Plan Mode, TodoWrite patterns | View |
Skills are designed for EVM-compatible chains and work out of the box with Foundry's multi-chain tooling:
Mainnet / L2s Testnets
────────────── ──────────────
Ethereum (ETH) Sepolia
Polygon (MATIC) Mumbai
Arbitrum One Arbitrum Sepolia
Optimism OP Sepolia
Base Base Sepolia
Avalanche C-Chain Fuji
BNB Smart Chain BSC Testnet
zkSync Era zkSync Sepolia
Linea Linea Sepolia
Pharos Network Pharos Testnet
Chain-specific foundry.toml configuration, RPC URL management, and cast commands are all covered in the deploy and debug skills.
Your Project
│
├── CLAUDE.md ◄── skill auto-invoke rules
│
└── .agents/
└── 0xlayerghost/
└── solidity-agent-kit/
├── solidity-coding/ ──► Enforces code style & NatSpec
├── solidity-security/ ──► Guards access control & reentrancy
├── solidity-testing/ ──► Fuzz + invariant test coverage
├── solidity-checklist/ ──► 6-layer on-chain op verification
├── solidity-deploy/ ──► Pre/post deploy checklist
├── solidity-debug/ ──► Cast-based tx analysis
├── solidity-audit/ ──► Manual security review
├── defi-security/ ──► DeFi-specific attack vectors
├── git-workflow/ ──► Conventional commits & PRs
└── claude-code-usage/ ──► Agent context strategies
AI Agent Workflow:
─────────────────────────────────────────────────────────────
User Request ──► CLAUDE.md detects task type
│
┌────────────┼────────────┐
▼ ▼ ▼
Writing .sol Testing .t.sol Deploying .s.sol
│ │ │
/solidity-coding /solidity- /solidity-deploy
/solidity-security testing /defi-security
│ │ │
└────────────┴────────────┘
│
forge build
forge test
│
✅ Verified
| Agent | Status | Notes |
|---|---|---|
| Claude Code | ✅ Full support | Best experience — CLAUDE.md auto-invoke |
| Cursor | ✅ Full support | .cursor/rules integration |
| Windsurf | ✅ Full support | .windsurf/rules integration |
| Trae | ✅ Supported | .trae integration |
| OpenCode | ✅ Supported | — |
| Gemini CLI | ✅ Supported | — |
| GitHub Copilot | 🚧 Partial | Manual invocation |
| Codex CLI | ✅ Supported | — |
| Tool | Version | Purpose |
|---|---|---|
| Solidity | ^0.8.20 |
Smart contract language |
| Foundry | latest | Build, test, deploy (forge, cast, anvil) |
| OpenZeppelin | 4.9.x |
Secure contract standards (ERC20/721/1155/Governor) |
| Slither | latest | Static analysis (90+ vulnerability detectors) |
| skills.sh | latest | Skill package management & distribution |
solidity-agent-kit/
├── CLAUDE.md.template ← Copy to project root as CLAUDE.md
├── skills/
│ ├── solidity-coding/ ← Coding standards & naming conventions
│ │ └── SKILL.md
│ ├── solidity-security/ ← Security best practices
│ │ └── SKILL.md
│ ├── solidity-testing/ ← Test organization & coverage
│ │ └── SKILL.md
│ ├── solidity-checklist/ ← 6-layer checklist for on-chain ops
│ │ └── SKILL.md
│ ├── solidity-deploy/ ← Deployment workflow & checks
│ │ └── SKILL.md
│ ├── solidity-debug/ ← On-chain transaction debugging
│ │ └── SKILL.md
│ ├── solidity-audit/ ← Manual security audit checklist
│ │ └── SKILL.md
│ ├── defi-security/ ← DeFi-specific attack prevention
│ │ └── SKILL.md
│ ├── git-workflow/ ← Git conventions for Solidity projects
│ │ └── SKILL.md
│ └── claude-code-usage/ ← AI agent context strategies
│ └── SKILL.md
├── openai/
├── LICENSE
└── README.md
- Solidity developers using Foundry (
forge,cast,anvil) - Teams building DeFi protocols, NFT projects, and DAO governance
- Developers using AI coding agents who want consistent, secure output
- Security engineers running automated audits with Slither + manual checklists
- Anyone deploying on EVM-compatible chains
MIT — see LICENSE
Built by 0xlayerghost
Blockchain & Solidity Engineer
If this helped you ship safer contracts, give it a ⭐