refactor(artifact-writer): rename evaluators to graders in grading.json

apps/cli/src/commands/eval/artifact-writer.ts

@@ -26,7 +26,7 @@ export interface GradingArtifact {
     readonly total_tool_calls: number;
     readonly errors_encountered: number;
   };
-  readonly evaluators?: readonly {
+  readonly graders?: readonly {
     readonly name: string;
     readonly type: string;
     readonly score: number;
@@ -219,12 +219,12 @@ function buildAssertions(result: EvaluationResult): GradingArtifact['assertions'
 }
 
 // ---------------------------------------------------------------------------
-// Build evaluators list
+// Build graders list
 // ---------------------------------------------------------------------------
 
 function buildEvaluators(
   scores: readonly EvaluatorResult[] | undefined,
-): GradingArtifact['evaluators'] {
+): GradingArtifact['graders'] {
   if (!scores || scores.length === 0) {
     return undefined;
   }
@@ -267,7 +267,7 @@ export function buildGradingArtifact(result: EvaluationResult): GradingArtifact
       total_tool_calls: totalToolCalls,
       errors_encountered: errorsEncountered,
     },
-    evaluators: buildEvaluators(result.scores),
+    graders: buildEvaluators(result.scores),
     workspace_changes: parseWorkspaceChanges(result.fileChanges),
     conversation: result.conversationId
       ? {

apps/cli/src/commands/pipeline/bench.ts

@@ -133,7 +133,7 @@ export const evalBenchCommand = command({
         assertions: allAssertions,
         summary: { passed, failed, total: allAssertions.length, pass_rate: passRate },
         execution_metrics: { tool_calls: {}, total_tool_calls: 0, errors_encountered: 0 },
-        evaluators: evaluators.map((e) => ({
+        graders: evaluators.map((e) => ({
           name: e.name,
           type: e.type,
           score: e.score,

apps/cli/src/commands/results/export.ts

@@ -7,7 +7,7 @@
  *     benchmark.json           — aggregate scores, pass/fail counts, timing
  *     index.jsonl              — per-test manifest with artifact pointers
  *     <test-id>/
- *       grading.json           — per-test grading artifact (assertions, evaluators)
+ *       grading.json           — per-test grading artifact (assertions, graders)
  *       timing.json            — per-test timing artifact
  *       outputs/
  *         response.md          — human-readable agent response for this test

apps/cli/src/commands/results/manifest.ts

@@ -135,7 +135,13 @@ function hydrateManifestRecord(baseDir: string, record: ResultManifestRecord): E
       evidence: assertion.evidence,
     })),
     scores:
-      grading?.evaluators?.map((evaluator) => ({
+      // `evaluators` was renamed to `graders` in v4.13 — read both for backwards compat with old artifacts.
+      // TODO: remove `evaluators` fallback once old run directories are no longer in use.
+      (
+        grading?.graders ??
+        (grading as (GradingArtifact & { evaluators?: GradingArtifact['graders'] }) | undefined)
+          ?.evaluators
+      )?.map((evaluator) => ({
         name: evaluator.name,
         type: evaluator.type,
         score: evaluator.score,

apps/cli/test/commands/eval/artifact-writer.test.ts

@@ -127,10 +127,10 @@ describe('buildGradingArtifact', () => {
 
     const grading = buildGradingArtifact(result);
 
-    expect(grading.evaluators).toHaveLength(2);
-    expect(grading.evaluators?.[0].name).toBe('format-check');
-    expect(grading.evaluators?.[0].type).toBe('code-grader');
-    expect(grading.evaluators?.[1].score).toBe(0.7);
+    expect(grading.graders).toHaveLength(2);
+    expect(grading.graders?.[0].name).toBe('format-check');
+    expect(grading.graders?.[0].type).toBe('code-grader');
+    expect(grading.graders?.[1].score).toBe(0.7);
   });
 
   it('records error as errors_encountered', () => {
@@ -150,7 +150,7 @@ describe('buildGradingArtifact', () => {
       total: 0,
       pass_rate: 0,
     });
-    expect(grading.evaluators).toBeUndefined();
+    expect(grading.graders).toBeUndefined();
   });
 
   it('includes workspace_changes when fileChanges present', () => {

apps/cli/test/commands/eval/pipeline/bench.test.ts

@@ -75,7 +75,7 @@ describe('pipeline bench', () => {
     const grading = JSON.parse(await readFile(join(OUT_DIR, 'test-01', 'grading.json'), 'utf8'));
     expect(grading.summary.pass_rate).toBeGreaterThan(0);
     expect(grading.assertions.length).toBeGreaterThan(0);
-    expect(grading.evaluators).toHaveLength(2);
+    expect(grading.graders).toHaveLength(2);
   });
 
   it('writes index.jsonl with one entry per test', async () => {

apps/cli/test/commands/eval/pipeline/pipeline-e2e.test.ts

@@ -49,7 +49,7 @@ describe('eval pipeline e2e', () => {
     const grading = JSON.parse(
       await readFile(join(OUT_DIR, 'input-test', 'test-01', 'grading.json'), 'utf8'),
     );
-    expect(grading.evaluators).toHaveLength(2);
+    expect(grading.graders).toHaveLength(2);
     expect(grading.summary.pass_rate).toBeGreaterThan(0);
 
     const indexContent = await readFile(join(OUT_DIR, 'index.jsonl'), 'utf8');

apps/cli/test/commands/results/export-e2e-providers.test.ts

@@ -472,8 +472,8 @@ describe('export e2e — multi-provider metrics verification', () => {
       expect(grading.execution_metrics.tool_calls.Write).toBe(1);
 
       // Evaluators
-      expect(grading.evaluators).toHaveLength(1);
-      expect(grading.evaluators?.[0].name).toBe('accuracy');
+      expect(grading.graders).toHaveLength(1);
+      expect(grading.graders?.[0].name).toBe('accuracy');
     });
 
     it('should produce correct grading for Copilot CLI result with mixed assertions', async () => {

apps/cli/test/commands/results/export.test.ts

@@ -261,10 +261,10 @@ describe('results export', () => {
     expect(grading.execution_metrics).toBeDefined();
 
     // Has evaluators
-    expect(grading.evaluators).toBeDefined();
-    expect(grading.evaluators).toHaveLength(1);
-    expect(grading.evaluators?.[0].name).toBe('greeting_quality');
-    expect(grading.evaluators?.[0].type).toBe('llm-grader');
+    expect(grading.graders).toBeDefined();
+    expect(grading.graders).toHaveLength(1);
+    expect(grading.graders?.[0].name).toBe('greeting_quality');
+    expect(grading.graders?.[0].type).toBe('llm-grader');
 
     const perTestTimingPath = path.join(artifactDir(outputDir, RESULT_FULL), 'timing.json');
     expect(existsSync(perTestTimingPath)).toBe(true);

plugins/agentv-dev/skills/agentv-bench/references/eval-yaml-spec.md

@@ -323,7 +323,7 @@ LLM grader results are read from disk at `<test-id>/llm_grader_results/<name>.js
 ```
 
 **Output:**
-- `<test-id>/grading.json` — merged grading with `evaluators`, `assertions`, `summary.pass_rate`
+- `<test-id>/grading.json` — merged grading with `graders`, `assertions`, `summary.pass_rate`
 - `index.jsonl` — one JSON line per test: `{test_id, score, pass, evaluators: [...]}`
 - `benchmark.json` — aggregate stats: `{metadata: {targets}, run_summary: {<target>: {mean, stddev, n}}}`