chore(web): address Copilot review on the tab smoke

Three findings, all valid.

**The read/get stages asserted an RPC status, not a render.** #2148 asks
for the resource content and the prompt messages to *render*, and
`data-read-status` / `data-get-status` flip the moment the call resolves
— so the smoke stayed green with the panel deleted. `ResourcePreviewPanel`
and `PromptMessagesDisplay` now carry `data-testid="resource-preview"` /
`"prompt-messages"`, and each stage waits on the status **and** the
rendered panel. Verified by breaking each testid: the smoke fails with
`reading `events` did not render a preview — last observed:
data-read-status="ok"`, which is the finding stated back exactly.

**`smoke:web:tabs` followed ambient `SMOKE_BROWSER`**, so
`SMOKE_BROWSER=firefox npm run ci` would have run the supposedly
Chromium-only smoke under Firefox — contradicting both this PR's stated
scope and the gated-tier invariant in `run-engine-smokes.test.mjs`. The
npm script now pins `chromium` on **both** halves (install and launch)
and the smoke resolves through `resolveRequestedBrowser(argv, env)`, where
an explicit argument wins over the environment; a new case in that same
test file pins it, including that the smoke stays out of `ENGINE_SMOKES`.

**The AGENTS rules were not mirrored into `.github/copilot-instructions.md`.**
Correct, and worse than an omission: that file's engine bullet asserts the
headless web smokes are "engine-parameterized, not Chromium-only", which
this PR made false. It now names `smoke:web:tabs` as the deliberate
exception, and carries the genuinely review-citable new rule — assert
`data-*`, never copy, and never stop at an RPC status where the ask was
that something render.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UL9retfvAXRgvi6EWk4SY4
Signed-off-by: cliffhall <cliff@futurescale.com>

scripts/run-engine-smokes.test.mjs

@@ -90,6 +90,23 @@ describe("every engine tier consumes ENGINE_SMOKES", () => {
     }
   });
 
+  it("the Chromium-only tab smoke pins its engine on BOTH halves", () => {
+    // `smoke:web:tabs` is not engine-sensitive and deliberately stays out of
+    // ENGINE_SMOKES — but "Chromium-only" is a claim the script has to keep.
+    // Without the literal engine it followed ambient `SMOKE_BROWSER`, so
+    // `SMOKE_BROWSER=firefox npm run ci` would have run it under Firefox
+    // (Copilot, #2148). Both halves are pinned because they must AGREE: an
+    // install of one engine and a launch of another surfaces as a confusing
+    // "Executable doesn't exist" rather than as the wiring error it is.
+    assert.match(
+      scripts["smoke:web:tabs"],
+      /install-smoke-browser\.mjs chromium .*smoke-web-tabs\.mjs chromium$/,
+    );
+    // And it must not be in ENGINE_SMOKES, or the Firefox tier would pick it
+    // up regardless of what its own script pins.
+    assert.ok(!ENGINE_SMOKES.includes("smoke-web-tabs.mjs"));
+  });
+
   it("each gated tier names its engine explicitly rather than reading the env", () => {
     // An ambient SMOKE_BROWSER must not be able to redirect a gate: without the
     // literal engine, `SMOKE_BROWSER=firefox npm run ci` would run Firefox twice

scripts/smoke-web-tabs.mjs

@@ -30,12 +30,17 @@
  *      only summarize it, so a result panel that renders text alone looks
  *      correct while dropping the data.
  *   2. **Resources** — the list renders, a resource is read, and the preview
- *      populates. Templates are asserted present (the RFC 6570 fixture, #1919),
- *      which is what proves `resources/templates/list` reached the screen —
- *      that list is fetched by a *different* call than `resources/list` and can
- *      fail on its own.
- *   3. **Prompts** — the list renders and `simple_prompt` is fetched (selecting
- *      an argument-less prompt auto-fetches), and the messages render.
+ *      panel renders its contents. Templates are asserted present (the RFC 6570
+ *      fixture, #1919), which is what proves `resources/templates/list` reached
+ *      the screen — that list is fetched by a *different* call than
+ *      `resources/list` and can fail on its own.
+ *   3. **Prompts** — the list renders, `simple_prompt` is fetched (selecting an
+ *      argument-less prompt auto-fetches), and the messages panel renders.
+ *
+ * The read/get stages assert the RPC status **and** the rendered panel, not the
+ * status alone. Each `data-*-status` flips the moment its call resolves, so a
+ * smoke stopping there would stay green with the panel deleted — which is the
+ * opposite of what a connected-flow smoke is for.
  *
  * Deliberately **one script over one browser launch** rather than one per tab:
  * launching the browser and booting the server dominate the cost, and
@@ -79,11 +84,8 @@
 import { resolve } from "node:path";
 import { startProdWebServer } from "./lib/prod-web-server.mjs";
 import { stopChild } from "./lib/child-cleanup.mjs";
-import {
-  attachPageDiagnostics,
-  loadBrowser,
-  resolveBrowserName,
-} from "./lib/headless-browser.mjs";
+import { attachPageDiagnostics, loadBrowser } from "./lib/headless-browser.mjs";
+import { resolveRequestedBrowser } from "./install-smoke-browser.mjs";
 import {
   buildConnectDeepLink,
   connectViaDeepLink,
@@ -92,14 +94,20 @@ import { startMcpAppServer } from "./lib/mcp-app-flow.mjs";
 
 const repoRoot = resolve(import.meta.dirname, "..");
 
-// Resolved before anything starts, so an unsupported SMOKE_BROWSER fails
-// immediately rather than after a web server and an MCP server are up. Every
-// message carries the engine so a failure names which run it came from — even
-// though this smoke is Chromium-only in every gate, it can still be pointed
-// elsewhere by hand.
+// The engine comes from an explicit ARGUMENT first, and only then from
+// `SMOKE_BROWSER`. That precedence is the load-bearing part: `smoke:web:tabs`
+// passes `chromium` to both this and the installer, so an ambient
+// `SMOKE_BROWSER=firefox npm run ci` cannot silently redirect a gated tier that
+// claims to be Chromium-only — the same invariant `run-engine-smokes.test.mjs`
+// pins for the engine tiers (Copilot, #2148). Running another engine by hand is
+// still one argument away: `node scripts/smoke-web-tabs.mjs firefox`.
+//
+// Resolved before anything starts, so an unsupported value fails immediately
+// rather than after a web server and an MCP server are up. Every message
+// carries the engine, so a failure names which run it came from.
 let BROWSER;
 try {
-  BROWSER = resolveBrowserName();
+  BROWSER = resolveRequestedBrowser(process.argv.slice(2), process.env);
 } catch (err) {
   console.error(
     `smoke:web:tabs FAILED — ${err instanceof Error ? err.message : String(err)}`,
@@ -279,10 +287,16 @@ try {
       .getByRole("button", { name: RESOURCE, exact: true })
       .click({ timeout: 30_000 });
     await waitForStage(page, {
-      selector: '[data-testid="resources-screen"][data-read-status="ok"]',
+      // Both the RPC status AND the rendered preview. The status flips the
+      // moment `resources/read` resolves, so stopping there would stay green
+      // with the preview panel removed entirely (Copilot, #2148) — and #2148
+      // asks for the content to *render*, not merely for the call to return.
+      selector:
+        '[data-testid="resources-screen"][data-read-status="ok"]' +
+        ' [data-testid="resource-preview"]',
       screen: '[data-testid="resources-screen"]',
       attrs: ["data-read-status"],
-      what: `reading \`${RESOURCE}\` did not succeed`,
+      what: `reading \`${RESOURCE}\` did not render a preview`,
     });
 
     // ── Prompts ────────────────────────────────────────────────────────────
@@ -299,10 +313,14 @@ try {
       .getByRole("button", { name: new RegExp(`^${PROMPT}`) })
       .click({ timeout: 30_000 });
     await waitForStage(page, {
-      selector: '[data-testid="prompts-screen"][data-get-status="ok"]',
+      // As with the resource preview: the messages panel must have rendered,
+      // not merely `prompts/get` returned (Copilot, #2148).
+      selector:
+        '[data-testid="prompts-screen"][data-get-status="ok"]' +
+        ' [data-testid="prompt-messages"]',
       screen: '[data-testid="prompts-screen"]',
       attrs: ["data-get-status"],
-      what: `getting \`${PROMPT}\` did not succeed`,
+      what: `getting \`${PROMPT}\` did not render its messages`,
     });
   };