Mercury

A Hermes Agent skill for multi-chain blockchain cash flow analysis, fraud detection, trading strategy generation, and real-time interactive dashboards.

Features

• Multi-Chain Support — Ethereum, Base, and Solana transaction fetching
• Cash Flow Analysis — Period-by-period inflow/outflow breakdown, token distribution, top counterparties
• Fraud Detection — 6 pattern detectors (dust attacks, rapid transfers, mixer interaction, address poisoning, dormant activation, high fan-out)
• Trading Strategy — AI-generated trading insights based on wallet behavior patterns
• Interactive Dashboard — Three-layer WebGL + Canvas + HTML dashboard with:
  • SpectraNoise-inspired neural network noise background
  • Live network particle animation with mouse interaction
  • v0-style stat cards, gauges, charts, and risk grid
• Hermes Subagents — Parallel execution with delegate_task for maximum speed

Installation

Quick Install (for Hermes Agent users)

# Clone into your Hermes skills directory
git clone https://github.com/YOUR_USERNAME/mercury.git ~/.hermes/skills/data-science/blockchain-cashflow

API Keys

Add your API keys to ~/.hermes/.env:

ETHERSCAN_API_KEY=your_etherscan_key    # Free at https://etherscan.io/apis
BASESCAN_API_KEY=your_basescan_key      # Free at https://basescan.org/apis
HELIUS_API_KEY=your_helius_key          # Free at https://helius.dev

The scripts auto-load keys from ~/.hermes/.env — no manual export needed.

Usage

With Hermes Agent (Recommended)

Just tell your Hermes agent:

Analyze wallet 0x8a4ff766a5dfb16d9dbd6a31f950c48a0caf0f54

The agent will automatically:

1. Detect the chain (Ethereum/Base/Solana)
2. Fetch all transactions using parallel subagents
3. Run cash flow analysis + fraud detection simultaneously
4. Generate a trading strategy
5. Launch the interactive dashboard

Standalone Scripts

# 1. Fetch transactions
python3 scripts/fetch_transactions.py \
  --address 0x8a4ff766a5dfb16d9dbd6a31f950c48a0caf0f54 \
  --chain ethereum \
  --include-tokens \
  --output /tmp/bc_txns.json

# 2. Analyze cash flow
python3 scripts/cashflow_analyzer.py \
  --input /tmp/bc_txns.json \
  --address 0x8a4ff766a5dfb16d9dbd6a31f950c48a0caf0f54 \
  --output /tmp/bc_cashflow.json

# 3. Detect fraud patterns
python3 scripts/fraud_detector.py \
  --input /tmp/bc_txns.json \
  --address 0x8a4ff766a5dfb16d9dbd6a31f950c48a0caf0f54 \
  --output /tmp/bc_fraud.json

# 4. Generate trading strategy
python3 scripts/trading_strategy.py \
  --cashflow /tmp/bc_cashflow.json \
  --fraud /tmp/bc_fraud.json \
  --output /tmp/bc_strategy.json

# 5. Launch dashboard
python3 scripts/dashboard_server.py \
  --cashflow /tmp/bc_cashflow.json \
  --fraud /tmp/bc_fraud.json

Architecture

Mercury Skill
├── SKILL.md                    # Hermes skill definition (YAML frontmatter + instructions)
├── scripts/
│   ├── fetch_transactions.py   # Multi-chain transaction fetcher (Etherscan V2 / Helius)
│   ├── cashflow_analyzer.py    # Period-by-period cash flow analysis
│   ├── fraud_detector.py       # 6-pattern fraud detection engine
│   ├── trading_strategy.py     # Trading strategy generator
│   ├── visualizer.py           # Terminal-based ASCII visualization
│   ├── dashboard_server.py     # Python HTTP server for the dashboard
│   └── dashboard.html          # Three-layer interactive dashboard
└── README.md

Dashboard Preview

The dashboard features three animation layers:

• Layer 0: WebGL fragment shader with fbm noise, domain warping, scanlines, and film grain
• Layer 1: Canvas 2D network particles with proximity connections and mouse attraction
• Layer 2: Dashboard UI with stat cards, flow network map, circular gauges, charts, and risk grid

Fraud Detection Patterns

Pattern	Severity	Description
Dust Attacks	Medium	Tiny unsolicited token transfers
Rapid Transfers	Medium	Burst activity suggesting bots or wash trading
Mixer Interaction	Critical	Transfers to/from known mixing services
Address Poisoning	High	Dust from addresses mimicking legitimate ones
Dormant Activation	Medium	Long-inactive wallet suddenly active
High Fan-Out	High	Single-block mass distribution

Requirements

• Python 3.8+
• No external dependencies (stdlib only!)
• Hermes Agent (for subagent orchestration)

License

MIT

Credits

Built for the Nous Research Hermes Agent Hackathon — March 2026