| name | sharpclaw |
| description | Autonomous Solana meme-coin trading bot. Use this skill when the user wants to trade tokens, check portfolio/positions/P&L, analyze tokens, manage watchlists, or control auto-trading. Triggers: buy/sell tokens, check balance, view positions, analyze sentiment, rug check, portfolio status, pause/resume trading, watchlist management.
|
| user-invocable | true |
| metadata | {"openclaw":{"emoji":"๐พ","requires":{"env":["WALLET_PRIVATE_KEY","SOLANA_RPC_URL"],"bins":["bun"]},"primaryEnv":"WALLET_PRIVATE_KEY"}} |
SharpClaw - Solana Meme-Coin Trading Bot
You control a running SharpClaw trading bot instance. The bot runs as a background systemd service (claw-trading.service) and executes trades autonomously. You interact with it by querying its SQLite database and using its CLI.
How to Send Commands
SharpClaw exposes a CLI for direct interaction. Run commands via:
cd ~/claw_trading && bun run src/cli.ts <command> [args...]
To query the database directly (for read-only operations):
cd ~/claw_trading && sqlite3 data/trading.db "<SQL query>"
Available Commands
Trading
Buy a token:
bun run src/cli.ts buy <mint_address> <sol_amount>
Example: bun run src/cli.ts buy So11111111111111111111111111111111111111112 0.1
Sell a token:
bun run src/cli.ts sell <mint_address> [token_amount]
Omit amount to sell entire position.
Portfolio & Status
Check wallet balance:
bun run src/cli.ts balance
Returns SOL balance and all token holdings.
View open positions:
bun run src/cli.ts positions
Shows each position with entry price, current price, P&L %, and stop-loss level.
Today's P&L:
bun run src/cli.ts pnl
Returns daily realized P&L, per-strategy expectancy, win rates, R:R ratios, and adaptive factor status (confidence scale, stop adjustments, auto-disable).
Recent trades:
bun run src/cli.ts trades
Bot status:
bun run src/cli.ts status
Shows: auto-trading on/off, circuit breaker state, total exposure, daily loss tracking.
Analysis
AI-powered token analysis:
bun run src/cli.ts analyze <mint_address>
Runs full pipeline: technical indicators (RSI, MACD, volume), social sentiment (via xAI Grok or Twitter), rug pull check, and AI recommendation with confidence score.
Watchlist
Add to watchlist:
bun run src/cli.ts watch <mint_address> [symbol]
Remove from watchlist:
bun run src/cli.ts unwatch <mint_address>
View watchlist:
bun run src/cli.ts watchlist
Auto-Trading Control
Pause auto-trading:
bun run src/cli.ts pause
Stops new entries. Existing positions continue to be managed (stop-loss/take-profit still active).
Resume auto-trading:
bun run src/cli.ts resume
Service Management
Check if the bot is running:
systemctl is-active claw-trading
View recent logs:
journalctl -u claw-trading --no-pager -n 30
Restart the bot:
systemctl restart claw-trading
Trading Philosophy
SharpClaw is designed for low win rate (~30%), high reward-to-risk ratio (4-5x). This is intentional:
- Frequent small losses are expected and normal
- The system profits from rare large gains (fat-tailed meme-coin returns)
- Circuit breaker triggers on cumulative P&L, not win/loss count
- Each strategy self-evolves: winning strategies get amplified, losing ones get attenuated or auto-disabled
Safety Rules
- Never bypass the circuit breaker - if it's triggered, wait for cooldown
- Always check status before manual trades - verify exposure limits
- Use analyze before manual buys - the rug check can save from scams
- Maximum position size is enforced automatically (default 0.5 SOL)
- The bot manages stop-losses automatically - don't manually sell positions that have active stops unless the user explicitly requests it
Interpreting Results
- Sentiment score: -100 (very bearish) to +100 (very bullish). Source shown as "Grok" (xAI) or "DexScreener" (synthetic)
- Hype score: 0-100 composite of volume, momentum, buy count, liquidity
- Rug score: 0-100 from RugCheck API. Lower is safer. >50 = high risk
- Strategy expectancy: positive = strategy has edge, negative = losing. Auto-disabled at z < -1.65 after 20+ trades
- Confidence scale: 0.3x-1.5x multiplier on position size based on historical performance
