diff --git a/.github/workflows/performance-monitor.yml b/.github/workflows/performance-monitor.yml
index db34cd61..9fc613dc 100644
--- a/.github/workflows/performance-monitor.yml
+++ b/.github/workflows/performance-monitor.yml
@@ -47,16 +47,29 @@ jobs:
       - name: Run benchmarks
         id: benchmark
         run: |
-          npx tsx scripts/ci/benchmark-performance.ts > benchmark-results.json 2>&1 || true
+          npx tsx scripts/ci/benchmark-performance.ts > benchmark-results.json 2>benchmark-progress.log || true
+          cat benchmark-progress.log
+          echo "--- JSON output ---"
           cat benchmark-results.json
 
       - name: Upload results
+        if: always()
         uses: actions/upload-artifact@v4
         with:
           name: benchmark-results-${{ github.sha }}
-          path: benchmark-results.json
+          path: |
+            benchmark-results.json
+            benchmark-progress.log
           retention-days: 90
 
+      - name: Validate benchmark JSON
+        run: |
+          if ! jq empty benchmark-results.json 2>/dev/null; then
+            echo "ERROR: benchmark-results.json is not valid JSON"
+            cat benchmark-results.json
+            exit 1
+          fi
+
       - name: Check for regressions
         id: check
         run: |
diff --git a/scripts/ci/benchmark-performance.ts b/scripts/ci/benchmark-performance.ts
index d0112cb6..4023ccfe 100644
--- a/scripts/ci/benchmark-performance.ts
+++ b/scripts/ci/benchmark-performance.ts
@@ -9,6 +9,10 @@
  *  - Docker network creation time
  *
  * Outputs structured JSON with mean, median, p95, p99 per metric.
+ *
+ * IMPORTANT: stdout/stderr contract:
+ *   - stdout (console.log): JSON result only — consumed by the CI workflow via redirect to file
+ *   - stderr (console.error): progress messages and diagnostics — kept separate so JSON stays valid
  */
 
 import { execSync, ExecSyncOptions } from "child_process";