Skill v1.0.0

currentLLM-judged scan90/100

kuaamu/swarmmind/swarm-orchestrate

──Details

PublishedOctober 22, 2026 at 11:27 AM

Content Hashsha256:c9339c3568701287...

Git SHAa12870b970ad

──Files

Files (1 file, 3.2 KB)

SKILL.md3.2 KBactive

SKILL.md · 86 lines · 3.2 KB

version: "1.0.0" name: swarm-orchestrate description: Run SwarmMind's autonomous DeFi intelligence cycle - coordinates market scanning, risk assessment, and trade execution across specialized AI agents with x402 micropayments on X Layer

SwarmMind Orchestration

Overview

SwarmMind's Portfolio Manager uses a ReAct (Reasoning + Acting) loop to autonomously coordinate a swarm of specialized DeFi agents. The LLM dynamically decides which agents to call, in what order, and how to react to results.

Quick Start

bash

# 1. Start all agents
cd packages/agents/alpha-scout && npm run dev &
cd packages/agents/risk-oracle && npm run dev &
cd packages/agents/trade-executor && npm run dev &
cd packages/agents/portfolio-manager && npm run dev &
 
# 2. Set a strategy
curl -X POST http://localhost:3000/strategy \
  -H "Content-Type: application/json" \
  -d '{"strategy": "Conservative DeFi trading on X Layer, max $50 per position"}'
 
# 3. Run one ReAct cycle
curl -X POST http://localhost:3000/orchestrate/react
 
# 4. View reasoning trace
curl http://localhost:3000/reasoning/latest

How the ReAct Loop Works

Context: Portfolio state + user strategy are provided to the LLM
Reason: LLM analyzes context and decides which tool to call
Act: Tool is executed (HTTP call to agent service with x402 payment)
Observe: Result is fed back to the LLM
Repeat: Until the LLM decides the cycle is complete

Available Tools

Tool	Agent	Description
`get_market_signals`	Alpha Scout	Fetch trading signals (ARBITRAGE, MOMENTUM, MEAN_REVERSION)
`assess_risk`	Risk Oracle	Evaluate signal risk (score 1-10, PROCEED/CAUTION/REJECT)
`assess_liquidity`	Liquidity Agent	Check pool depth (DEEP/ADEQUATE/SHALLOW/AVOID)
`execute_trade`	Trade Executor	Execute swap via OKX DEX aggregator
`get_portfolio_state`	Portfolio Manager	Current positions, balances, recent trades
`check_agent_status`	All	Health check for any agent

WebSocket Events

Connect to ws://localhost:3000 to receive real-time events:

TOOL_CALL - Tool invocation (request/response phases)
SIGNAL_DETECTED - New trading signal found
RISK_ASSESSED - Risk evaluation completed
TRADE_EXECUTED - Trade confirmed
LIQUIDITY_ASSESSED - Pool liquidity checked
PORTFOLIO_UPDATE - State updated after cycle

Strategy Configuration

Strategies are parsed by the LLM into structured parameters:

json

{
  "riskTolerance": "LOW | MEDIUM | HIGH",
  "maxPositionSize": 100,
  "preferredTokens": ["OKB", "USDC"],
  "strategyType": "CONSERVATIVE | BALANCED | AGGRESSIVE",
  "constraints": ["Only trade OKB/USDC pair"]
}

onchainos Cross-Skill Workflows

SwarmMind integrates with the onchainos skill ecosystem:

dex-signal + swarm-signals: Combine onchainos DEX signals with Alpha Scout analysis for broader market coverage
dex-swap + swarm-trade: Use onchainos swap routing alongside SwarmMind's risk-assessed execution
wallet-portfolio + swarm-orchestrate: Extend onchainos portfolio view with SwarmMind's autonomous management
onchain-gateway + swarm-orchestrate: Use onchainos gateway for cross-chain signals feeding into SwarmMind's ReAct loop

All versions