TrustTrade - Autonomous Kraken Trading Agent
TrustTrade is a fully autonomous crypto trading agent combining Kraken CLI execution with ERC-8004 on-chain identity, validation, and reputation. **Strategy v3 — Pro Trader Logic:** Trend-following strategy using EMA 20/50 crossover, RSI(14), and MACD(12,26,9) on live 1-hour candles across BTC, ETH, and SOL. Only buys in confirmed uptrends (EMA20 > EMA50). Manages exits with 5 layers: hard stop-loss (-1.5%), take-profit (+2%), trailing stop (locks profits at +0.8%), trend reversal exit, and RSI overbought exit (>70). In downtrends, stays in cash unless extreme reversal appears. **Risk Firewall — Deterministic Safety:** Before any order reaches Kraken, a deterministic policy engine validates: pair whitelist (BTC/ETH/SOL only), position sizing (max 6%, capped at $500 by RiskRouter), per-pair rate limiting (60s min), and a 2% daily loss circuit breaker. Every approved trade gets an EIP-712 TradeIntent signature. Sells are exempt from position sizing. **Full ERC-8004 Integration:** Agent ID #78 on Ethereum Sepolia, validation score 99. Every trade flows through the shared RiskRouter for on-chain approval (142 submissions, 77 approved). Filled trades post checkpoints to ValidationRegistry (708 on-chain). 5,000+ logged events including 1,319 firewall blocks — every decision is auditable. **Live Dashboard:** Real-time dashboard shows live Kraken prices (5s refresh), strategy signals with RSI/MACD, trade history with intent hashes, firewall blocks, RiskRouter events with Etherscan links, and on-chain checkpoints. **Safety by Design:** Paper-first development, dead man's switch, persistent cost basis tracking, 5-minute sell cooldown, and peak PnL tracking for trailing stops.
Category tags:
.png&w=96&q=75){kind=small}
"1. Application of Technology: 4 / 5 Justification: Excellent Python backend engineering. The strategy code dynamically calculates moving average crossovers via numpy and pandas, implements dynamic trailing stops based on peak PnL, and properly hits the Kraken CLI via subprocess. Connecting this raw math execution to an EIP-712 ERC-8004 signer satisfies all technical criteria perfectly. 2. Presentation: 4 / 5 Justification: Very strong. Giving the bot a live dashboard (milanp93.github.io/trading-agent/dashboard/) that tracks 142 on-chain validations and 1,319 firewall blocks provides incredible visibility into the agent's behavior. The presentation PDF drives home the "Safety by Design" philosophy. 3. Business Value: 4 / 5 Justification: High value as a retail trading tool. The focus on a "Dead Man's switch," 5-minute sell cooldowns, and trailing stop losses (locking in profits at +0.8%) makes this a highly pragmatic bot that a user could actually leave running unmonitored without losing their shirt. 4. Originality: 2.5 / 5 Justification: While beautifully executed, this project completely misses the "AI" component of the "AI Trading Agents" hackathon. There are zero LLM API calls, no machine learning components, and no generative reasoning. It is a traditional quantitative python bot (SMA crossover script) wired up to an ERC-8004 smart contract. ⚖️ Pros & Cons Pros: Pragmatic Execution: By ignoring LLM hallucinations and using standard mathematics (EMA crossovers and RSI > 70) to execute trades, the bot is significantly faster, cheaper to run, and less prone to catastrophic logic failures than LLM-driven competitors. Deep Risk Management: The 5-layer exit strategy (hard stop-loss, take-profit, trailing stop, trend reversal, RSI overbought) is exceptionally robust. Most entries only implement a simple static stop-loss. Full ERC-8004 Compliance: Every trade intent is cryptographically signed and approved by the on-chain RiskRouter before executing, meaning it fully qualifies for the Web3 Trustless track. Cons: Not an "AI" Agent: The code (strategy.py) reveals this is entirely a deterministic script using Pandas dataframes. There is no Generative AI (OpenAI, Claude, Groq) or Predictive AI (Scikit, PyTorch) present. It is a crypto trading script, not an AI agent. Opaque Execution Environment: Developing localized python scripts that write to local checkpoints.jsonl files means setting it up requires the user to manage their own Python environments, cron jobs, and .env variables, making it harder for non-technical users to deploy. "
Sanem Avcil
For Innovators & Creators