Function createEvaluationNextAction

plugins/plugin-eval/src/core/presentation.js:49–94

(result)

}

export function createEvaluationNextAction(result) {
  const labelName = targetLabel(result.target);

  if (hasStructuralFailures(result)) {
    return {
      label: "Fix structural issues first",
      why: "Failing manifest or skill structure issues reduce trust and can invalidate later measurements.",
      command: buildStartCommand(result.target, "What should I fix first?"),
      chatPrompt: "What should I fix first?",
    };
  }

  if (!result.observedUsage?.sampleCount && hasBudgetPressure(result)) {
    return {
      label: "Measure real token usage next",
      why: "The static budget looks heavy, so live usage is the fastest way to confirm whether the cost is acceptable.",
      command: buildStartCommand(result.target, `Measure the real token usage of this ${labelName}.`),
      chatPrompt: `Measure the real token usage of this ${labelName}.`,
    };
  }

  if (result.observedUsage?.sampleCount) {
    return {
      label: "Review the measurement plan",
      why: "You already have observed usage, so the highest-value next step is deciding what to instrument or improve.",
      command: buildStartCommand(result.target, "What should I run next?"),
      chatPrompt: "What should I run next?",
    };
  }

  if (hasWarningsOrFailures(result)) {
    return {
      label: "Fix the top findings and rerun the report",
      why: "A short pass on the highest-value findings will improve trust, readability, and the signal quality of future benchmarks.",
      command: buildStartCommand(result.target, "What should I fix first?"),
      chatPrompt: "What should I fix first?",
    };
  }

  return {
    label: "Choose the next workflow from chat",
    why: "The report is clean enough that the best next step depends on whether you want budgets, benchmarks, or comparisons.",
    command: buildStartCommand(result.target, "What should I run next?"),
    chatPrompt: "What should I run next?",
  };
}

export function applyEvaluationPresentation(result) {
  result.summary = enrichSummary(result, result.summary);