AFlow Harness Recovery Lead

Use this skill only when the aflow engine invokes it for harness recovery. Inspect the failure evidence, choose one supported recovery action, and return machine-readable JSON only.

Rules

• Read the provided failure evidence first, then decide whether the failure is repeatable or a one-off.
• Return exactly one JSON object. Do not wrap it in markdown, prose, or code fences.
• The JSON object must contain these keys:
  • action
  • delay_seconds
  • reason
  • suggested_keywords
  • suggested_action
• action must be one of:
  • retry_same_team_after_delay
  • switch_to_backup_team_and_retry
  • fail_immediately
• delay_seconds must be an integer or null.
• reason should be short and specific.
• suggested_keywords should contain reusable phrases that could become future deterministic recovery rules.
• suggested_action may repeat the chosen action or be null.
• Do not return extra keys.
• Do not return prose-only output.

What To Optimize For

• Prefer retry_same_team_after_delay when the same team is likely to succeed after waiting.
• Prefer switch_to_backup_team_and_retry when a backup team is the safest next attempt.
• Prefer fail_immediately when the failure is not safely recoverable.