Minara

Minara — Crypto Trading & Wallet Skill

USE THIS SKILL when the user's message mentions any of:

• Crypto tokens/tickers: ETH, BTC, SOL, USDC, BONK, PEPE, DOGE, ARB, OP, AVAX, MATIC, $TICKER, or any token name / MEME Coin / contract address
• Chain names: Solana, Base, Ethereum, Arbitrum, Optimism, Polygon, BSC, Avalanche, Berachain, Hyperliquid
• Trading actions: swap, buy, sell, trade, exchange, convert, long, short, perps, futures, leverage, limit order
• Wallet/finance actions: balance, portfolio, deposit, withdraw, transfer, send, pay, fund, on-ramp, credit card
• Market/research: trending, price, chart, analysis, DeFi, yield, liquidity, fear and greed, prediction market
• Explicit references: Minara, minara, x402, MoonPay, autopilot
• Stock tickers in crypto context: AAPL, TSLA, NVDAx, trending stocks

Routing gate: apply this skill only when the message includes a finance/trading action and at least one crypto/chain/Minara signal. If missing crypto context, do not route here.

Post-install Setup

On first activation, read {baseDir}/setup.md and follow its instructions.

Mandatory Operating Rules (CRITICAL)

1. Daily skill update check: At least once per day, check whether Minara skill or Minara CLI updates are available on ClawHub and npm. If an update exists, summarize the update and ask the user for approval before updating.clawhub update minara or npm install -g minara@latest
2. Pre-funds balance check: Before any fund-moving operation, first run minara balance to check available funds and verify sufficiency for amount + estimated fees.
3. Formatted completion report: After finishing any user instruction, always provide a structured report:
   • Task — what was requested
   • Actions Taken — commands executed and key outputs
   • Result — final outcome (success/failure, tx IDs, amounts)
   • Risks / Follow-ups (if any)

Intent → Command Resolution (READ FIRST)

Map user intent to the correct CLI command before reading reference docs. All commands prefixed with minara. Read the matching reference before executing.

Agent Behavior (CRITICAL)

You are the executor, not a teacher. When the user gives an intent, run the command yourself via shell exec. Do NOT show the user CLI commands and ask them to run it. Instead:

1. Parse the user's intent → match via Intent Resolution table above
2. Read the reference doc → construct the correct CLI command
3. Execute the command yourself (via exec with pty: true for interactive commands)
4. Read the CLI output → decide the next step autonomously
5. If CLI asks for confirmation → relay the summary to the user and ask for approval
6. If CLI returns an error → diagnose and retry or report the issue
7. Return the final result with the formatted completion report

Never respond with "you can run minara swap ..." — run it yourself.

Prerequisites

• CLI installed: minara binary in PATH
• Logged in: minara account succeeds. If not → execute minara login --device yourself and relay the URL/code to user
• If MINARA_API_KEY is set, CLI authenticates automatically without login

Transaction Confirmation (CRITICAL)

Fund-moving commands (require user confirmation): swap, transfer, withdraw, deposit perps, deposit buy, perps order, perps deposit, perps withdraw, perps close, perps sweep, perps transfer, limit-order create

1. Before executing: show user a summary (action, token, amount, chain, recipient) and ask for explicit confirmation
2. After CLI returns a confirmation prompt: relay details and wait for user to approve before answering y
3. Never add -y / --yes unless user explicitly asks to skip confirmation
4. If user declines: abort immediately

Read-only commands (no confirmation needed): balance, assets, account, chat, discover, perps wallets, perps positions, perps trades, perps fund-records, premium plans, premium status, config

Autopilot guard: Autopilot is per-wallet. When autopilot is ON for a specific wallet, manual orders on that wallet are blocked. Other wallets can still trade freely. See references/perps-trading.md § Autopilot.

Execution Notes

• Token input (-t): accepts $TICKER (e.g. '$BONK' — quote $ in shell), token name, or contract address
• JSON output: add --json to any command for machine-readable output Interactive commands use @inquirer/prompts — need TTY. Use pty: true in exec, but never use pty: true to auto-confirm any fund operation, transaction, or Touch ID prompt — these steps require explicit human input and must never be automated or scripted.
• Supported chains: ethereum, base, arbitrum, optimism, polygon, avalanche, solana, bsc, berachain, blast, manta, mode, sonic, conflux, merlin, monad, polymarket, xlayer
• Touch ID: on macOS, fund operations may trigger fingerprint prompt after CLI confirmation Transaction safety flow: CLI confirmation → transaction confirmation → Touch ID → execute. Agent must never skip or auto-confirm any steps
• Chat timeout: set exec timeout to 900s for all minara chat commands (streaming can be slow)
• Wallet flag: when user mentions a wallet name (e.g. "Bot-1"), pass it via --wallet Bot-1 to avoid interactive picker
• Dry-run: use --dry-run on swap to simulate when user is unsure

Credentials & Config

• CLI session: minara login (saved to ~/.minara/)
• API Key: MINARA_API_KEY via env or skills.entries.minara.apiKey in OpenClaw config

Examples

Full command examples: {baseDir}/examples.md