Skip to main content

Notes

  • The score is read from the task’s output (the agent self-scores). This makes gating synchronous with task completion, with no engine change required.
  • For independent verification, attach an async LLM-judge scorer via the <Task scorers={...}> prop. That score lands in _smithers_scorers for telemetry and is not consulted for gating.
  • currentScore = null | undefined means “pending or missing.” Default behavior is onPending: "continue": the loop runs another iteration and the next task’s output should populate the score.
  • <LoopUntilScored> is a thin wrapper around <Loop>. Caller resolves the score from ctx; the component only computes until and forwards.