fix(ci): require WASM cache for Windows SEA builds

Windows builds cannot reliably download ML models from Hugging Face due to network issues. Fail fast with clear error message directing users to run build-wasm.yml workflow first to prime the cache.

This prevents 30-60 minute build failures and makes the dependency explicit.

Author: jdalton

.github/workflows/build-sea.yml

@@ -203,6 +203,13 @@ jobs:
           key: node-sea-${{ matrix.platform }}-${{ matrix.arch }}-${{ steps.sea-cache-key.outputs.hash }}
           restore-keys: node-sea-${{ matrix.platform }}-${{ matrix.arch }}-
 
+      - name: Check WASM cache requirements (Windows)
+        if: matrix.os == 'windows' && (steps.yoga-cache.outputs.cache-hit != 'true' || steps.ai-cache.outputs.cache-hit != 'true' || steps.onnx-cache.outputs.cache-hit != 'true')
+        run: |
+          echo "::error::WASM assets not cached for Windows build. Run build-wasm.yml workflow first to prime the cache."
+          echo "Windows builds require pre-cached WASM assets due to network reliability issues."
+          exit 1
+
       - name: Build CLI (required for SEA)
         if: steps.sea-cache.outputs.cache-hit != 'true' || inputs.force
         run: pnpm --filter @socketsecurity/cli run build