@plaited/agent-eval-harness 0.12.1 → 0.12.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@plaited/agent-eval-harness",
-  "version": "0.12.1",
+  "version": "0.12.2",
   "description": "CLI tool for capturing agent trajectories from headless CLI agents",
   "license": "ISC",
   "engines": {
@@ -56,12 +56,12 @@
     ]
   },
   "dependencies": {
-    "@plaited/development-skills": "0.7.0",
+    "@plaited/development-skills": "0.8.0",
     "zod": "^4.3.6"
   },
   "devDependencies": {
-    "@biomejs/biome": "2.3.12",
-    "@types/bun": "1.3.6",
+    "@biomejs/biome": "2.3.14",
+    "@types/bun": "1.3.9",
     "format-package": "7.0.0",
     "lint-staged": "16.2.7",
     "typescript": "5.9.3"

package/src/pipeline/compare-trials.ts

@@ -262,6 +262,7 @@ const computeTrialsQualityMetrics = (results: TrialResult[]): QualityComputeResu
 
   return {
     metrics: {
+      type: 'trial',
       avgScore: sum / rawScores.length,
       medianScore: percentile(sorted, 0.5),
       p25Score: percentile(sorted, 0.25),

package/src/pipeline/compare.ts

@@ -365,6 +365,7 @@ export const runCompare = async (config: ExtendedCompareConfig): Promise<Compari
     const fails = results.length - passes
 
     quality[label] = {
+      type: 'run',
       avgScore: scores.length > 0 ? scores.reduce((a, b) => a + b, 0) / scores.length : 0,
       passRate: results.length > 0 ? passes / results.length : 0,
       passCount: passes,

package/src/pipeline/tests/compare-statistical.spec.ts

@@ -105,6 +105,10 @@ describe('runCompare statistical strategy', () => {
     // Verify reliability metrics include type discriminator
     expect(report.reliability.high?.type).toBe('run')
     expect(report.reliability.low?.type).toBe('run')
+
+    // Verify quality metrics include type discriminator
+    expect(report.quality.high?.type).toBe('run')
+    expect(report.quality.low?.type).toBe('run')
   })
 
   test('computes confidence intervals for performance metrics', async () => {

package/src/pipeline/tests/compare-trials.spec.ts

@@ -477,6 +477,7 @@ describe('runTrialsCompare', () => {
     expect(report.quality?.run1).toBeDefined()
 
     const qual = report.quality?.run1
+    expect(qual?.type).toBe('trial')
     expect(qual?.avgScore).toBeGreaterThan(0)
     expect(qual?.medianScore).toBeGreaterThan(0)
     expect(qual?.p25Score).toBeDefined()

package/src/schemas/schemas.ts

@@ -620,6 +620,8 @@ export type QualityConfidenceIntervals = z.infer<typeof QualityConfidenceInterva
  * Quality metrics for a single run in comparison.
  */
 export const QualityMetricsSchema = z.object({
+  /** Discriminator for run-level quality metrics */
+  type: z.literal('run'),
   /** Mean grader score (0-1) */
   avgScore: z.number(),
   /** Percentage of pass=true results */
@@ -942,6 +944,8 @@ export type TrialsQualityConfidenceIntervals = z.infer<typeof TrialsQualityConfi
  * Only present when a grader was used during trials capture.
  */
 export const TrialsQualityMetricsSchema = z.object({
+  /** Discriminator for trial-level quality metrics */
+  type: z.literal('trial'),
   /** Average score across all trials */
   avgScore: z.number(),
   /** Median score */