| name | prompt-strategy-setup |
| shortcut | prompt |
| description | This skill should be used when the user explicitly asks to CREATE or BUILD an AI-decision (Prompt-based) trading strategy. Trigger phrases include "create a prompt strategy", "build an AI trading strategy", "set up AI decision trading", "help me create a prompt-based strategy". Do NOT trigger for general questions about how strategies work or what prompt strategies are. |
| description_zh | 当用户要求创建或构建 AI 决策(Prompt)交易策略时使用此技能,引导完成完整的策略搭建流程。 |
Prompt Strategy Setup (AI Decision)
Guide the user through creating a complete AI-decision trading pipeline
using Trading Prompts. This strategy type uses LLM interpretation for
trade decisions — best for complex judgment, sentiment, and context.
Pre-requisites (MUST confirm before proceeding)
- Confirm the target exchange (Hyperliquid or Binance)
- Confirm the environment (Testnet or Mainnet)
- These determine which wallets, signal pools, and data sources are available
Workflow
Phase 1: Requirements Gathering
Understand the user's trading intent:
- Which symbol(s) to trade (e.g., BTC, ETH, SOL)
- Trading style (trend following, mean reversion, breakout, etc.)
- Risk parameters (leverage, position size, stop loss preferences)
- Trigger preference (signal-based, scheduled, or both)
Use the user's profile (trading_style, risk_preference, experience_level)
to pre-fill defaults where possible. Skip questions the profile already answers.
→ [CHECKPOINT] Summarize understood requirements. Wait for user confirmation.
Phase 2: Signal Pool Configuration
Query existing signal pools: list_signal_pools
- Filter by the confirmed exchange — signal pool exchange MUST match
- If a suitable pool exists, propose reusing it
- If no match, delegate to Signal AI:
call_signal_ai
- Include exchange, symbol, desired trigger frequency in the task
- Save the result:
save_signal_pool
A Prompt Trader can bind MULTIPLE signal pools. Suggest combinations
if the strategy benefits from multi-signal triggers.
→ [CHECKPOINT] Show signal pool(s) to be used. Wait for user confirmation.
Phase 3: Trading Prompt Creation
Delegate to Prompt AI: call_prompt_ai
- Include: symbol, trading style, risk parameters, exchange context
- Prompt AI will use proper variables ({current_price}, {market_regime}, etc.)
Save the result: save_prompt
→ [CHECKPOINT] Show prompt summary. Wait for user confirmation.
Phase 4: AI Trader Assembly
Query existing traders: list_traders
- If a suitable trader exists (correct exchange, has wallet), propose reusing
- If not, create one:
create_ai_trader (LLM config only)
Bind the prompt: bind_prompt_to_trader(trader_id, prompt_id)
- This is ONE-TO-ONE: replaces any existing prompt binding
Configure triggers: update_trader_strategy(trader_id, signal_pool_ids, ...)
- signal_pool_ids: array of pool IDs (can be multiple)
- scheduled_trigger_enabled: true/false
- trigger_interval: seconds between scheduled triggers
→ [CHECKPOINT] Show complete configuration summary:
- Trader name and LLM model
- Bound prompt name
- Signal pool(s) with conditions
- Trigger settings
Wait for user confirmation.
Phase 5: Activation Guide
These are SECURITY OPERATIONS that require manual user action:
- Bind Wallet → AI Trader → click the trader → bind wallet section
- Hyperliquid: create API Wallet on Hyperliquid website, paste agent private key + master wallet address
- Binance: paste API key + secret key
- Wallet exchange must match the confirmed exchange
- Start Trading → AI Trader → click the trader → "Start Trading" toggle
Offer to verify after user completes: diagnose_trader_issues(trader_id)
Key Rules
- Signal pool exchange MUST match the trader's wallet exchange
- Never guess thresholds — always delegate to Signal AI
- Never write prompts yourself — always delegate to Prompt AI
- One Trader = One Prompt (one-to-one binding)
- One Trader = Multiple Signal Pools (one-to-many trigger config)