moss-site/moss-trade-bot-skills

Moss – AI‑Powered Crypto Trading Bot Factory

What it is

• A Python‑based framework that lets you describe a crypto‑trading style in plain English (e.g., “trend‑following, low leverage, breakout”) and automatically creates a quantitative trading agent.
• The system builds the strategy, runs a local back‑test on historic Hyperliquid data, and then iteratively evolves the parameters week‑by‑week based on the back‑test results.
• Designed for research/education – it is not financial advice.

Key components

Component	What it does
Natural‑Language to Strategy	Parses a user‑provided description, infers technical‑indicator choices, leverage, risk limits, and creates a JSON‑style parameter file.
Back‑testing engine	Runs cross‑margin simulations on 15‑minute Hyperliquid CSV datasets, reporting Sharpe, max‑drawdown, win‑rate, etc.
Weekly Evolution Loop	After each back‑test segment the AI applies seven “Reflection Principles” (e.g., analyse wins/losses, micro‑adjust parameters within ±30 %) to produce a new tuned parameter set while keeping the overall “personality” (bias, leverage, risk) fixed.
Safety guardrails	Enforces leverage caps, mandatory wide stop‑losses for high‑leverage trades, and confirmation steps before live deployment.
Platform integration (optional)	Allows binding to the Moss web platform, uploading back‑test results for verification, and launching a live‑trading bot that talks to the Hyperliquid exchange.

How to get started

1. Clone the pinned release (v1.0.28) and install Python 3 dependencies (pandas, numpy, ccxt).
2. Download the data cache – the first run automatically pulls a tarball of Hyperliquid 15‑minute CSVs into ~/.cache/moss‑trade‑bot‑factory/....
3. Create a fingerprint for the dataset you’ll use (e.g., BTC/USDC) – this records the local CSV path.
4. Run a standard back‑test or the recommended evolution back‑test which segments the data (default 672 bars ≈ 7 days) and performs the reflection‑adjust cycle.
5. (Optional) Bind to the Moss platform with a pair‑code, upload your evolved results, and start live trading via live_trade.py and live_runner.py.

Typical workflow example

# 1. Clone and install
git clone --depth 1 --branch v1.0.28 https://github.com/moss-site/moss-trade-bot-skills.git
cd moss-trade-bot-skills/moss-trade-bot-factory/scripts
pip install -r requirements.txt

# 2. Fetch a dataset fingerprint (first run downloads the CSV)
python3 fetch_data.py \
  --data data_cache/hyperliquid_BTCUSDC_15m_2025-07-01_304d.csv \
  --symbol BTC/USDC \
  --timeframe 15m > /tmp/fingerprint.json

# 3. Run the evolution back‑test (weekly adaptation)
python3 run_evolve_backtest.py \
  --data $(jq -r .csv_path /tmp/fingerprint.json) \
  --params-file /tmp/bot_params.json \
  --segment-bars 672 \
  --capital 10000 \
  --output /tmp/evolve_result.json

Why it matters

• Bridges the gap between non‑technical trading ideas and executable quantitative strategies using LLM‑driven parameter inference.
• Iterative self‑tuning mimics a research loop: back‑test → analyse → tweak → repeat, all automated.
• Local‑first execution keeps data and logic on your machine; the optional cloud platform is merely a verification and live‑deployment layer.

Community & licensing

• Open‑source under the permissive MIT‑0 license.
• Active Discord, Telegram, and Twitter channels for support and leaderboard of community‑created agents.

---

All details above are taken directly from the repository’s README; no additional features have been inferred.