web3-agent-skills
vaquarkhan/web3-agent-skills
Production-grade Agent Skills for Web3 AI agents: 15 workflows, 6 MCP servers, DeFi/NFT/compliance guardrails, and VS Code/JetBrains installers.
Instalação
npx convez-ai vaquarkhan/web3-agent-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
Web3 Agent Skills
GitHub Releases · VS Code Extension · JetBrains Plugin
Production-grade Agent Skills for Web3 AI agents: 15 domain workflows, 6 MCP servers, chain presets, compliance guardrails, MEV protection, and VS Code / JetBrains installers.
The open skill registry and execution toolkit for Web3 agents — wallets, DeFi, NFTs, DAOs, cross-chain, compliance, and institutional RWA workflows with the same discipline used by production Web3 teams.
Agent Skills Registry Compatibility
- Every capability lives in a directory containing a
SKILL.md - Every
SKILL.mdstarts with YAML frontmatter:nameanddescription - Supporting materials in
references/,knowledge-base/,examples/,hooks/, andscripts/ - Progressive disclosure: agents load skills on demand
Quick Start
git clone https://github.com/vaquarkhan/web3-agent-skills.git
cd web3-agent-skills
npm install && npm run build
node scripts/link-skills.js
npm test
Install into your project:
scripts/install.sh --tool all --target /path/to/project
# Windows:
pwsh scripts/install.ps1 --tool all --target C:\path\to\project
Bootstrap shortcut:
./bootstrap.sh /path/to/project
pwsh .\bootstrap.ps1 C:\path\to\project
Install Surfaces
VS Code / Cursor / Windsurf / VSCodium
- Download
.vsixfrom GitHub Releases - Command Palette → Extensions: Install from VSIX...
- Command Palette → Web3 Agent Skills: Install Full Toolkit
See vscode-extension/README.md and tutorials/installing-vscode-and-jetbrains-plugins.md.
JetBrains (IntelliJ, PyCharm, WebStorm, GoLand)
- Download
.zipfrom Releases or install from JetBrains Marketplace - Settings → Plugins → Install from Disk...
- Tools → Web3 Agent Skills → Install Full Toolkit
See jetbrains-plugin/README.md and docs/jetbrains-setup.md.
Install By Tool
| Tool | Setup guide | Install path |
|---|---|---|
| VS Code / Cursor | docs/cursor-setup.md | Extension or scripts/install.sh --tool cursor |
| JetBrains | docs/jetbrains-setup.md | Plugin or scripts/install.sh --tool all |
| Claude | docs/claude-setup.md | .claude/commands/ or install script |
| Copilot | docs/copilot-setup.md | .github/copilot-instructions.md |
| Codex / AGENTS.md | docs/getting-started.md | AGENTS.md + skills-index.md |
Core Principles
- Simulate before every state-changing transaction
- Screen all addresses against sanctions lists
- Never expose private keys or seed phrases
- Contract verification before interaction
- Exact token approvals over unlimited
- MEV protection for high-value swaps
- $10,000 confirmation threshold
Lifecycle Commands
| Command | Purpose |
|---|---|
/simulate | Simulate transaction (required before send) |
/audit | Security audit and rug-pull checks |
/screen | Sanctions screening and Travel Rule |
/swap | DeFi swap with MEV + slippage guards |
/bridge | Cross-chain transfer with security review |
/review | Gas, compliance, and security review |
15 Skills
| # | Skill | Phase | Capabilities |
|---|---|---|---|
| 1 | Wallet Management | 1 | HD wallets, ERC-4337, multi-sig, social recovery |
| 2 | Smart Contract Interaction | 1 | ABI, deploy, proxies, multicall |
| 3 | DeFi Operations | 1 | Swap, lend/borrow, stake, LP, flash loans |
| 4 | NFT Management | 2 | Mint/trade, ERC-721/1155, TBA |
| 5 | DAO Governance | 4 | Proposals, voting, treasury, timelocks |
| 6 | Cross-Chain Bridge | 3 | Bridging, messaging, intent routing |
| 7 | Token Analytics | 2 | Whale tracking, DEX volume, PnL |
| 8 | Security Audit | 2 | Vulnerability detection, rug pulls |
| 9 | Compliance & AML | 3 | Sanctions, Travel Rule, MiCA, Howey |
| 10 | Identity (DID) | 3 | DIDs, ENS, ZK-KYC, attestations |
| 11 | Oracle & Data Feeds | 4 | Chainlink, Pyth, VRF, keepers |
| 12 | Gas Optimization | 1 | EIP-1559, nonce, private txs, blobs |
| 13 | RWA Tokenization | 4 | Treasuries, ERC-3643, yield |
| 14 | Social & Reputation | 5 | Lens, Farcaster, reputation |
| 15 | MEV Protection | 2 | Flashbots, TWAP, sandwich protection |
Full catalog: skills-index.md
6 MCP Servers
| Server | Purpose |
|---|---|
| blockchain-rpc-server | Chain RPC (Alchemy, Infura, QuickNode) |
| defi-protocol-server | Protocol ABIs, subgraphs, pools |
| nft-metadata-server | IPFS, Arweave, NFT APIs |
| price-feed-server | Chainlink, Pyth, CoinGecko |
| compliance-screening-server | Chainalysis, sanctions, Travel Rule |
| ipfs-storage-server | Pinata, web3.storage |
Templates: mcp/ · Build: npm run build
Examples
| Example | Type | Best for |
|---|---|---|
| uniswap-swap-guardrails | Runnable | Safe DEX swap with simulation |
| compliance-screening-transfer | Runnable | Sanctions + Travel Rule |
| nft-mint-ipfs | Runnable | IPFS metadata + CID validation |
| defi-yield-vault-blueprint | Blueprint | ERC-4626 vault architecture |
| cross-chain-bridge-blueprint | Blueprint | L1→L2 bridge plan |
| dao-governance-blueprint | Blueprint | Governor + timelock |
See examples/README.md.
Project Structure
web3-agent-skills/
├── skills/ # 15 Agent Skills (SKILL.md per domain)
├── tools/mcp-servers/ # 6 MCP tool servers
├── knowledge-base/ # Chains, protocols, standards, regulations
├── guardrails/ # Web3 safety rules
├── agents/ # Multi-skill orchestration
├── presets/ # Chain-specific operating profiles
├── starter-packs/ # Problem-area bundles
├── examples/ # Runnable scaffolds + blueprints
├── mcp/ # MCP config templates
├── vscode-extension/ # VS Code / Cursor installer
├── jetbrains-plugin/ # JetBrains IDE installer
├── registry/ # Machine-readable asset index
├── docs/ # Setup and onboarding guides
├── scripts/ # Install and validation scripts
└── tests/ # Coverage tests
Supported Chains
Ethereum, Solana, Polygon, Arbitrum, Base, Optimism, Avalanche, BNB, Cosmos, Bitcoin L2s, TON, Sui/Aptos
Per-chain details: knowledge-base/chains/
Guardrails
guardrails/web3-guardrails.yaml:
- Never expose private keys or seed phrases
- Block unverified/scam contracts
- Warn on unlimited token approvals
- Screen addresses against sanctions lists
- Require simulation before execution
- Max value without confirmation: $10,000 USD
Multi-Agent Packaging
.cursor/rules/— Cursor rules.claude/commands/— Claude slash commands.github/copilot-instructions.md— GitHub CopilotAGENTS.md— Generic agent entry pointCLAUDE.md— Claude-specific notes
Related Projects
- data-engineering-agent-skills — same installer pattern for data engineering
- Agent Skills standard
Contributing
See CONTRIBUTING.md · CHANGELOG.md · SECURITY.md
License
MIT — see LICENSE