How does the skill handle transaction security and user funds?

Safety is managed through mandatory fork-chain simulations before any signature request, ensuring the agent sees the outcome before the user signs. Because it is non-custodial, private keys never leave the user's wallet.

Which blockchain networks and DeFi protocols are supported?

The skill is designed for versatility, supporting any EVM-compatible chain and providing ready-to-use tools for major protocols like Uniswap, Aave, Lido, Polymarket, and GMX.

What components are included in this skill package?

You receive the Aomi CLI, access to the Aomi runtime, over 40 pre-tuned protocol applications, and the full SDK suite (NPM/Rust) to integrate on-chain execution into your own agent UI or MCP.

Can this skill be used to interact with custom or unlisted smart contracts?

Yes, it features low-level primitives like `encode_and_call` and `stage_tx` that allow it to interact with any arbitrary smart contract, even if it isn't one of the 40+ pre-tuned protocols.

What AI platforms or agent frameworks can I use with aomi-transact?

The skill is compatible with Claude/Cowork plugins, blockchain Model Context Protocols (MCPs), and any custom AI agent built using the Aomi SDK or web3 assistant frameworks.

FreemarketingUniversal

aomi-transact

Build AI-driven crypto agents that simulate and execute transactions on any EVM chain via natural language.

skill install https://www.promptspace.in/skills/aomi-transact

Use this skill to build natural-language crypto agents, web3 assistants, trading bots, blockchain MCPs, or Claude/Cowork plugins that read and write EVM chain state. Aomi turns user prompts ("swap 1 ETH for USDC", "open a 3x GMX long", "bet $100 yes on this Polymarket question") into wallet-signed transactions on Ethereum, Base, Arbitrum, Optimism, Polygon, and Linea — generic, non-custodial, with fork-chain simulation before every signature. Trigger when the user wants to: scaffold a crypto/DeFi agent, build an AI trading or wallet assistant, wrap an EVM protocol as MCP tools, create natural-language interfaces to APIs from Across / 1inch / GMX / Hyperliquid / Polymarket / Binance / OKX, or add on-chain execution to universal protocols (e.g. Uniswap / Aave / Lido / Morpho) to an agent stack. Aomi CLI connects to Aomi runtime reads and writes EVM chain state: stages calldata, simulates as a batch on a forked chain, and returns wallet requests for the user to sign. Works on any EVM contract via low-level primitives: ABI encoding (encode_and_call), fork-batch simulation (simulate_batch), staged-tx queueing (stage_tx), and wallet handoff with EIP-712 (commit_tx, commit_eip712) — including multi-step swap-approve-execute routing. Ships with 40+ tuned protocol apps and all on-chain smart contracts. Every action simulates first, batches when possible, and is signed by the end-user wallet — keys never leave the user. --- ## About Aomi Aomi Labs builds native harness around blockchains functioning like Claude Code on-chain. We specialize in executions against arbitrary protocol with non-custodial workflow, account abstraction, and full security with simulations. Aomi also host agentic applications deployed and owned by developers, companies, and agents. Aomi provides E2E integration with UI, Skills and SDKs. **Links:** - 🌐 Website: https://aomi.dev - 🤖 Agents: https://aomi.dev/agents - 𝕏 Twitter: https://x.com/aomi_labs - 💻 GitHub: https://github.com/aomi-labs - 📦 Packages: - https://www.npmjs.com/package/@aomi-labs/widget-lib - https://www.npmjs.com/package/@aomi-labs/client - https://crates.io/crates/aomi-sdk --- ## Security OWASP AST03 permission manifest in SKILL.md frontmatter (risk_tier: L2 — signs/broadcasts on-chain transactions). Per-control AST01–AST10 walkthrough in SECURITY.md (included in this zip). Captured scanner reports (in source repo at https://github.com/aomi-labs/skills/tree/main/.scanner-reports): - Cisco AI Defense skill-scanner: PASS, 0 findings - pors/skill-audit: PASS, 0 errors - NMitchem/SkillScan: PASS, Risk 2.0/10 - Snyk agent-scan: PASS (advisory) — 4 W-code characterizations of the intentional risk surface, all documented with mitigations

Use cases

Convert natural language prompts into staged blockchain transactions.
Simulate multi-step DeFi batches on a fork to prevent failed transactions.
Execute gasless transactions using integrated Account Abstraction.
Build on-chain trading bots or AI assistants for Uniswap, Aave, and GMX.
Inspect blockchain state and balances using conversational queries.

Example

Prompt

Swap 1 ETH for USDC on Base and simulate the transaction before signing.

Output

$ aomi chat "swap 1 USDC for WETH on Uniswap V3, send to my wallet" \
    --public-key 0xUserAddress --chain 1 --new-session

I've staged your USDC → WETH swap on Uniswap V3 (0.05% fee tier).

Transaction Batch:
  1. Approve Uniswap Router to spend USDC
  2. Swap 1 USDC → WETH on V3 0.05% pool, recipient = your wallet

⚡ Wallet request queued: tx-1
   to:    0xA0b86991c6218b36c1d19D4a2e9Eb0cE3606eB48 (USDC)
   value: 0
   chain: 1
⚡ Wallet request queued: tx-2
   to:    0x68b3465833fb72A70ecDF485E0e4C7bD8665Fc45 (Uniswap SwapRouter02)
   value: 0
   chain: 1

$ aomi tx simulate tx-1 tx-2

Simulation result:
  Batch success: true
  Stateful: true
  Total gas: 197194
  Step 1 — Approve Uniswap Router to spend USDC    success: true   gas: 55798
  Step 2 — Swap 1 USDC for WETH on Uniswap V3      success: true   gas: 141396

$ aomi tx sign tx-1 tx-2

Exec:    aa (alchemy, 7702)
✅ Sent! Hash: 0xc7f8...3a91

Known limitations

Requires @aomi-labs/client v0.1.30+ (older versions lack --aa flags and the simulation gate). Account-abstraction sponsorship on L2s (Base, Arbitrum, Optimism) is not guaranteed by the zero-config Alchemy proxy in v0.1.30 — users may need to fund the EOA with a small amount of native gas on the destination chain or configure a BYOK provider (Alchemy with Gas Manager policy or Pimlico). Apps requiring provider tokens (Binance, Polymarket, Dune, Neynar, etc.) need user-supplied credentials via `aomi secret add` — the skill never sets credentials on its own initiative. Drain-vector annotations (recipient/onBehalfOf/mintRecipient ≠ msg.sender) are guard-blocks, not bypassable — by design, calldata that would route funds to a non-self address fails simulation. Public RPCs may rate-limit (429) or fail auth (401); production signing benefits from a dedicated chain-matching RPC.

Frequently asked questions

aomi-transact enables AI agents to interpret natural language prompts and convert them into executable, simulated EVM transactions for tasks like swapping, lending, or betting across major chains like Ethereum, Base, and Arbitrum.