test/unit/bundle-sdk-patches.test.ts (1)

151-182: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Hardcoded snapchatPixel config duplicates registry.ts; risks drift.

This inline proxyConfigs.snapchatPixel mirrors the actual registry entry (packages/script/src/registry.ts) but isn't sourced from it, unlike the analogous suite in test/unit/rewrite-ast.test.ts which calls getProxyConfigs(). If the registry's sdkPatches/domains for snapchatPixel change, this test will keep passing against a stale snapshot instead of failing.

♻️ Suggested refactor: derive from registry

-        proxyConfigs: {
-          snapchatPixel: {
-            domains: ['sc-static.net', 'tr.snapchat.com', 'pixel.tapad.com'],
-            sdkPatches: [
-              { type: 'replace-new-url-host', host: 'sc-static.net' },
-              { type: 'replace-script-loader-url', fromDomain: 'tr.snapchat.com', pathPrefix: '/config' },
-            ],
-          } as any,
-        },
+        proxyConfigs: {
+          snapchatPixel: (await buildProxyConfigsFromRegistry(await registry())).snapchatPixel as any,
+        },

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@test/unit/bundle-sdk-patches.test.ts` around lines 151 - 182, The Snapchat
proxy test is hardcoding a stale `proxyConfigs.snapchatPixel` snapshot instead
of using the shared registry data. Update the `applies snapchat config host
patches to bundled proxy scripts` test to derive `proxyConfigs` from the
registry source used elsewhere (for example via the same helper pattern as
`rewrite-ast.test.ts` and `getProxyConfigs()`), so `sdkPatches` and `domains`
stay aligned with `registry.ts` and future registry changes are caught.

test/unit/transform.test.ts (1)

645-687: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value

Drop the hash mock or assert the cache key. vi.mocked(hash).mockImplementationOnce(() => 'scevent.min') only affects the intermediate source filename/cache key; the emitted asset URL still comes from the fetched content hash. If this test isn’t checking cache naming, the mock is unnecessary.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@test/unit/transform.test.ts` around lines 645 - 687, The `transform` test for
`useScriptSnapchatPixel` is over-mocking `hash` in a way that only affects the
intermediate cache key, not the emitted asset URL. Remove the
`vi.mocked(hash).mockImplementationOnce(...)` stub unless the test explicitly
asserts cache-key behavior, or update the expectations to verify the cache key
produced by the `bundle` path and its interaction with `mockBundleStorage`. Keep
the assertions focused on the actual `transform` output and the fetch bypass
behavior.

test/fixtures/first-party/nuxt.config.ts (1)

4-5: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value

Grammatically broken comment.

"Pages that need the real SDK load provide an explicit trigger" doesn't parse. Likely meant "Pages that need the real SDK should provide an explicit trigger...".

✏️ Suggested wording fix

 // trigger: 'manual' prevents the auto-generated plugin from loading all 18
-// scripts globally on every page. Pages that need the real SDK load provide an
-// explicit trigger in their composable call, keeping cross-provider noise low.
+// scripts globally on every page. Pages that need the real SDK should provide
+// an explicit trigger in their composable call, keeping cross-provider noise low.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@test/fixtures/first-party/nuxt.config.ts` around lines 4 - 5, The inline
comment in the Nuxt config is grammatically broken and should be rewritten for
clarity. Update the descriptive text near the existing explanatory comment so
that the sentence about pages needing the real SDK reads naturally, using the
same surrounding context in the configuration block. Keep the meaning intact
while fixing the wording in the comment text itself.