AI agents are getting smarter.
Now they need the skills to actually do things.
Velvet Unicorn now gives builders and agents access to 1,600+ ranked skills across 25 categories, creating one of the largest open libraries of agent skills anywhere, with a major focus on blockchain.
The goal is simple:
Turn AI that knows into AI that does.
Try Now: https://vu.velvetdao.xyz/skill
Give Your Agent Onchain Superpowers
Velvet Unicorn Skills gives agents access to capabilities across nearly every major part of crypto:
Swaps & DEXs
Uniswap, Jupiter, aggregated routing across EVM and Solana.
Perps & Derivatives
Leverage, position sizing, liquidation calculations, and order routing.
DeFi & Yield
Lending, borrowing, liquidity, staking, liquid staking, and farming across protocols such as Aave and Lido.
Centralized Exchanges
Binance, OKX, Kraken, KuCoin, Bitget, Gate, and more.
Cross-Chain
Bridging, route discovery, fee quotes, and cross-chain execution.
Wallets & Portfolios
Balances, transfers, portfolio analysis, and non-custodial signing.
Analytics & Research
Onchain metrics, holder analysis, token screening, wallet activity, and market research.
Security
Honeypot and rug detection, contract risk analysis, and MEV-protected transactions.
Plus skills spanning oracles, NFTs, prediction markets, payments, identity, L1/L2 tooling, Solidity security, testing, and blockchain infrastructure.
Ranked, Not Just Indexed
Having 1,600+ skills isn’t useful if your agent can’t find the right one.
That’s why every skill is indexed by its name, tags, and full text, then ranked.
When an agent needs to accomplish something, it can find the best skill for the job, rather than simply using the first result it finds.
This turns Skills from a directory into a capability layer agents can actually navigate.
More Than Skills
Velvet Unicorn also provides the intelligence and infrastructure needed to connect these capabilities into real workflows.
Through API and MCP support, builders can plug Velvet Unicorn directly into their own agents, bots, dashboards, and applications.
VU can combine:
Market data
Technical analysis
Social sentiment
News
Onchain signals
Wallet context
Risk parameters
This allows agents to move beyond generic market analysis and understand the actual context surrounding a user, wallet, or strategy.
From Research to Execution
Velvet Unicorn isn’t designed to stop at analysis.
An agent could receive an instruction like:
“Swap 1,000 USDC into ETH on Base with a maximum 1% slippage.”
Velvet’s execution infrastructure can turn that intent into quote and transaction data, including routing, expected output, gas estimates, simulation results, and approvals before the wallet authorizes the final transaction.
The same infrastructure enables cross-chain trading, removing the need for users to manually bridge before executing on another network.
Prompt-to-Strategy
Skills can also become part of longer-running autonomous workflows.
With Prompt-to-Strategy, a natural-language idea can become a structured trading strategy.
For example:
“Build a BTC momentum strategy using RSI, MACD, moving averages, and volume confirmation with defined entries, exits, and risk controls.”
Or:
“Build a Pump.fun memecoin strategy that monitors new launches, wallets, liquidity, volume, and social momentum to identify opportunities.”
Describe the strategy.
Let the agent build the workflow.
Built for Agents
Velvet Unicorn can also operate beyond traditional chat interfaces.
VU can be integrated into Python automation, shell scripts, backend services, cron jobs, and other agent workflows.
With x402 integration, agents can even discover and pay for DeFi intelligence and services programmatically.
The result is an infrastructure layer built not just for humans using AI, but for agents interacting with other agents and financial systems autonomously.
How to use it
No SDK, no account. Point any agent (or yourself) at it:
npx -y https://vu.velvetdao.xyz/skill/unicorn-skills.tgz find “swap 0.1 ETH to USDC on Base”You get ranked JSON back instantly. Want the full recipe (instructions/body) for the top hits? Add --detail full. Prefer readable output? Add --pretty.
On-chain queries that just work:
find "open a 5x long on ETH"→ the best perps venues + how each sizes and manages the positionfind "is this token a scam or honeypot?"→ security skills that vet liquidity, taxes, owner privileges, holders — before funds movefind "best stablecoin yield right now"→ Aave & lending skills, ranked, with how to supply and read the APY/health factorfind "bridge USDC from Ethereum to Base"→ bridge skills with route + fee quotesfind "swap tokens on Solana"→ Jupiter and Solana DEX skillsfind "analyze this wallet's risk and PnL"→ wallet-analysis skillsfind "send a private transaction"→ MEV-protected routing to dodge sandwich attacks
Wire find into your agent’s runtime, feed the top skill’s instructions into the model’s context, and let it execute. That’s a full on-chain operator assembled from skills instead of hand-built one protocol at a time.
From AI That Knows to AI That Does
The first generation of AI answered questions.
The next generation takes action.
1,600+ ranked skills.
25 categories.
API + MCP.
Onchain intelligence.
Cross-chain execution.
Prompt-to-Strategy.
Research. Analyze. Decide. Act.
Velvet Unicorn is building the capability layer for onchain agents.
Try Now: https://vu.velvetdao.xyz/skill