hammergpt Performance Review
Analyze trading performance and provide actionable optimization suggestions. Combines Attribution AI analysis with memory-based historical context.
Pre-requisites (MUST confirm before proceeding)
- Confirm which trader or strategy to analyze
- Confirm the time period (last 7 days, 30 days, specific range)
- Confirm the exchange and environment
Workflow
Phase 1: Performance Data Collection
list_traders(trader_id)→ get trader details and current statusget_wallet_status(wallet_address)→ current balance and positions
Delegate deep analysis to Attribution AI:
call_attribution_ai(task="Analyze trading performance for trader X over the last N days. Identify patterns in winning and losing trades.")
→ [CHECKPOINT] Present performance summary in plain language:
- Overall P&L
- Win rate and profit factor
- Best and worst trades
- Common patterns in losses Wait for user to ask questions or request optimization.
Phase 2: Insight Extraction
From the analysis, identify:
- Strategy strengths (what's working)
- Strategy weaknesses (what's not)
- Market conditions where strategy underperforms
- Risk management observations
These insights will be automatically saved to user memory by the context compression system for future reference.
→ [CHECKPOINT] Present key insights and ask if user wants optimization suggestions.
Phase 3: Optimization Suggestions (if requested)
Based on findings, suggest specific improvements:
Signal Pool Adjustments:
- Trigger frequency too high/low
- Missing market regime filters
- Thresholds need recalibration
Strategy Logic Adjustments:
- Risk parameters (leverage, position size, stop loss)
- Entry/exit conditions
- Market regime awareness
If user agrees to optimize:
- Delegate to appropriate sub-agent with specific improvement instructions
- Use existing resource IDs (edit, not create new)
- Follow resource-management patterns
→ [CHECKPOINT] Show proposed changes before applying. Wait for user confirmation.
Key Rules
- Always delegate analysis to Attribution AI — don't guess performance data
- Present numbers in user-friendly format (percentages, not raw decimals)
- Be honest about poor performance — don't sugarcoat
- Always frame suggestions as options, not directives
- Remind users that past performance doesn't guarantee future results