diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2ac996e1..72b569ad 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -8,7 +8,22 @@ Semver applies from 1.0.0. A breaking change to a documented API needs a major 
 
 ## [Unreleased]
 
-Nothing yet.
+### Fixed
+
+- **A gate step whose suite executed nothing reports as skipped, never as passed** (#434). `x verify`
+  printed `✓ e2e 46ms` and `"skipped": false` over the one `e2eTest` a scaffold writes, which
+  `test.skip`s itself until the app registers a browser driver — so a customer's green gate included
+  a lane that had never run, and nothing in the table or in `--json` said so. `bun test` exits 0 over
+  an all-skipped file, but it prints its own counts and `packages/cli/src/test-counts.ts` was already
+  reading them: the rule is now `tests.ran === 0` in `packages/cli/src/verify-run.ts`, read once and
+  answered two ways — a step the committed `x.verify.json` requires stays the `X_VERIFY_SUITE_VANISHED`
+  failure it already was, and a step no floor requires is a skip beside `roadmap`'s, counted apart
+  from the passes in the summary line. A missing browser driver is a skip rather than an error code
+  because the framework ships no browser by design (`packages/scraping/src/cdp-port.ts`: the app
+  installs `puppeteer-core`, the CLI asks for it) — unlike Bun's own floor, which is `X_BUN_VERSION`
+  and a refusal. `StepResult.tests` now reaches `--json` and the human line, so `- e2e  found 1
+  test(s) and every one skipped itself` is distinguishable from `- roadmap`, which has no suite at
+  all.
 
 ## 20.1.0 - 2026-09-12
 
diff --git a/packages/cli/src/cmd-verify.test.ts b/packages/cli/src/cmd-verify.test.ts
index 1421a421..5ef8cd08 100644
--- a/packages/cli/src/cmd-verify.test.ts
+++ b/packages/cli/src/cmd-verify.test.ts
@@ -216,12 +216,15 @@ describe('unit · x verify', () => {
       });
     });
 
-    // The floor is the claim; without one there is nothing to be measured against, and a repo that
-    // never committed one is not ratcheted in either direction.
-    test('with no floor, an all-skipped suite stays the honest pass it was', async () => {
+    // The floor is the claim; without one there is nothing to be measured against, so nothing is
+    // FAILED here. It is still not a pass: a suite that executed nothing is the same "nothing to
+    // check" a step with no files reports, and issue #434 is what a green check over it costs.
+    test('with no floor, an all-skipped suite is a skip rather than a pass', async () => {
       await withFloor(undefined, async (root) => {
         const result = await runVerify([...green, allSkipped], { ...ctx, root });
         expect(result.ok).toBe(true);
+        expect(result.data).toMatchObject({ failed: [], skipped: ['live'] });
+        expect(result.steps?.find((step) => step.name === 'live')?.skipped).toBe(true);
       });
     });
 
diff --git a/packages/cli/src/e2e-driver.ts b/packages/cli/src/e2e-driver.ts
index ac9e559f..a16e089f 100644
--- a/packages/cli/src/e2e-driver.ts
+++ b/packages/cli/src/e2e-driver.ts
@@ -70,7 +70,8 @@ export const e2eFixtures = (page: PageLike, browser: E2eBrowserPage): E2eFixture
  * declaration `driverFixtures()` registered — the ordinary way a driver arrives, last registration
  * wins — so every `test('…', async ({ page }) => …)` in the suite gets a browser. `useE2eDriver`
  * is the other half: it is what makes `hasE2eDriver()` answer true and stops `e2eTest` becoming a
- * `test.skip` a green gate reports over.
+ * `test.skip` — which the gate now reports as a SKIPPED step rather than the green check it
+ * printed until #434, and which a repo whose `x.verify.json` names `e2e` gets red for.
  *
  * `budget`, `signIn` and `deploy` are deliberately NOT registered here. Each needs something a
  * page cannot supply — byte counts off a built `dist/`, an app's own sign-in route, a second build
diff --git a/packages/cli/src/messages.ts b/packages/cli/src/messages.ts
index 0c064d22..edd711e2 100644
--- a/packages/cli/src/messages.ts
+++ b/packages/cli/src/messages.ts
@@ -279,6 +279,14 @@ const CATALOG = {
   // The `errors` step's own coverage, in `output`: a scan without a parser reads most fix lines
   // and not all of them, and a step that reports findings alone claims a completeness it lacks.
   'cli.verify.fixCoverage': 'checked {checked} fix line(s), could not read {unreadable}',
+  // Why a skipped step is a dash when the step can say so: the suite exists and not one of its
+  // tests ran. Each test's own name carries the reason it skipped, and `bun test` prints only the
+  // counts — so the line says how many and `x test <type>` prints the names.
+  'cli.verify.allSkipped': 'found {skipped} test(s) and every one skipped itself',
+  // The other shape the same counts can take: a suite that reported no test at all, skipped ones
+  // included. Its own line and not `allSkipped` with a zero, because the repair differs — there is
+  // no skipped test to read a reason off, so the files the step selected hold no test.
+  'cli.verify.ranNothing': 'found no test to run in the file(s) it selected',
   'cli.verify.serial': 'serial',
   'cli.verify.workers': '{workers} workers',
   'cli.env.checked': '{count} declared variable(s), all present and valid',
diff --git a/packages/cli/src/output.test.ts b/packages/cli/src/output.test.ts
index 9ffaca28..5cd3ab3d 100644
--- a/packages/cli/src/output.test.ts
+++ b/packages/cli/src/output.test.ts
@@ -199,6 +199,28 @@ describe('unit · renderHuman says how a step was run', () => {
     expect(skipped).toContain('- unit');
   });
 
+  // #434: the whole point of reporting a zero-run suite as a skip is that a reader can tell it
+  // from a step with no suite at all, and the two shapes the counts take are two different
+  // repairs — read a skipped test's name, or find out why the files hold no test.
+  test('a skipped step that ran a suite says which kind of skip it is', () => {
+    const line = (tests: StepResult['tests']): string =>
+      renderHuman({
+        ok: true,
+        command: 'verify',
+        summary: 'ok',
+        steps: [
+          stepResult({ name: 'e2e', skipped: true, ...(tests === undefined ? {} : { tests }) }),
+        ],
+      });
+    expect(line({ ran: 0, skipped: 1 })).toContain(msg('cli.verify.allSkipped', { skipped: 1 }));
+    expect(line({ ran: 0, skipped: 0 })).toContain(msg('cli.verify.ranNothing'));
+    expect(line({ ran: 0, skipped: 0 })).not.toContain(
+      msg('cli.verify.allSkipped', { skipped: 0 }),
+    );
+    // A step with no suite — `roadmap` — has nothing to say beyond the dash.
+    expect(line(undefined).split('\n')[0]?.trim()).toBe('- e2e                40ms');
+  });
+
   test('a step with no workers field says nothing either', () => {
     expect(
       renderHuman({ ok: true, command: 'verify', summary: 'ok', steps: [stepResult({})] }),
diff --git a/packages/cli/src/output.ts b/packages/cli/src/output.ts
index 4d33c124..169f138f 100644
--- a/packages/cli/src/output.ts
+++ b/packages/cli/src/output.ts
@@ -4,6 +4,7 @@
 
 import { ERROR_DOCS_URL, renderThrowable, singleLine, stringField } from '@ultimat3/core';
 import { msg } from './messages';
+import type { TestCounts } from './test-counts';
 
 export interface Finding {
   readonly code: string;
@@ -24,6 +25,13 @@ export interface StepResult {
   readonly output?: string;
   /** Worker processes the step used; `1` means it ran serially. Absent for a non-test step. */
   readonly workers?: number;
+  /**
+   * What the step's suite executed. Absent for a step that spawned no test process, which is NOT
+   * the same state as `{ ran: 0 }` — a step with no suite and a suite whose every test skipped
+   * itself both report `skipped`, and this is the only thing that tells a reader which one it is
+   * looking at (#434).
+   */
+  readonly tests?: TestCounts;
 }
 
 export type JsonValue =
@@ -164,10 +172,27 @@ const width = (step: StepResult): string => {
   return `  ${step.workers === 1 ? msg('cli.verify.serial') : msg('cli.verify.workers', { workers: step.workers })}`;
 };
 
+/**
+ * Why a step is a dash, when the step itself can say. `- roadmap` is "there is nothing here to
+ * check"; a step that spawned a suite and executed none of it is a different state, which is what
+ * issue #434 reported as a green check — so a step carrying counts always says which it is, in one
+ * of the two shapes its counts can take. The reason each test skipped lives in that test's own
+ * NAME — `bun test` prints only the counts — so the line points at the suite and `x test <type>`
+ * is what prints the names.
+ */
+const why = (step: StepResult): string => {
+  const tests = step.tests;
+  if (step.skipped !== true || tests === undefined) return '';
+  if (tests.skipped === 0) return `  ${msg('cli.verify.ranNothing')}`;
+  return `  ${msg('cli.verify.allSkipped', { skipped: tests.skipped })}`;
+};
+
 export function renderHuman(result: CommandResult, verbose = false): string {
   const out: string[] = [];
   for (const step of result.steps ?? []) {
-    out.push(`  ${mark(step)} ${step.name.padEnd(18)} ${step.durationMs}ms${width(step)}`);
+    out.push(
+      `  ${mark(step)} ${step.name.padEnd(18)} ${step.durationMs}ms${width(step)}${why(step)}`,
+    );
     for (const finding of step.findings) out.push(renderFinding(finding, '      '));
     // NOT escaped, and that is the one exception: `output` is this process's own captured
     // subprocess stdout — `bun test`'s colour is the reason a human reads it at all, and it is
@@ -196,6 +221,9 @@ export function renderJson(result: CommandResult): string {
     skipped: step.skipped === true,
     findings: step.findings,
     ...(step.workers === undefined ? {} : { workers: step.workers }),
+    // The counts the human line's `why()` renders, as numbers: a `--json` reader deciding whether
+    // a skipped lane is missing a suite or missing a prerequisite needs the same fact CI's log has.
+    ...(step.tests === undefined ? {} : { tests: step.tests }),
     // A FAILED step carries its captured stdout, exactly as the human renderer prints it. CI runs
     // `--json`, and without this the log said only "one or more unit tests failed" with a generic
     // fix line — the failing test's name and its assertion diff existed and were thrown away, so
diff --git a/packages/cli/src/templates/scaffold-repo.ts b/packages/cli/src/templates/scaffold-repo.ts
index 61c19ed6..b58b820b 100644
--- a/packages/cli/src/templates/scaffold-repo.ts
+++ b/packages/cli/src/templates/scaffold-repo.ts
@@ -295,7 +295,9 @@ const biome = (): string => `{
  *
  * Two remain absent. `e2e` has a scaffolded file and it is an `e2eTest` — `test.skip` until the app
  * registers a browser driver, so the step would run zero tests and fail the ratchet on the
- * scaffold's own placeholder. `contract-diff` needs a committed `x.manifest.json`, which
+ * scaffold's own placeholder. Its absence is visible rather than silent: a suite that executed
+ * nothing reports as a SKIPPED step, not the green check it printed until #434, so the app's own
+ * gate line says `- e2e` and names it in the summary until a driver makes it run. `contract-diff` needs a committed `x.manifest.json`, which
  * `x manifest` writes later. Each joins the list in the commit that makes the app's gate run it.
  */
 const SCAFFOLD_FLOOR: readonly VerifyStepName[] = [
diff --git a/packages/cli/src/verify-run.test.ts b/packages/cli/src/verify-run.test.ts
index 262b4e74..93006f09 100644
--- a/packages/cli/src/verify-run.test.ts
+++ b/packages/cli/src/verify-run.test.ts
@@ -10,10 +10,11 @@ import { tmpdir } from 'node:os';
 // why: Bun exposes no path-join primitive; Bun.file and import() take one already joined.
 import { join } from 'node:path';
 import { msg } from './messages';
-import { exitCodeFor } from './output';
+import { exitCodeFor, renderHuman, renderJson } from './output';
 import { VERIFY_FLOOR_FILE } from './verify-floor';
 import { runVerify } from './verify-run';
 import type { VerifyContext, VerifyStep } from './verify-step';
+import { resetTestDiscovery, TEST_STEPS } from './verify-tests';
 
 /** The banner a narrowed run carries, from the catalog that renders it — never a second literal. */
 const NOT_A_GATE_RUN = msg('cli.verify.notAGateRun', { summary: '' }).trim();
@@ -169,3 +170,84 @@ describe('skips are counted apart from passes, and named', () => {
     expect(result.data).toMatchObject({ failed: ['drift'], skipped: ['e2e'] });
   });
 });
+
+// Issue #434: `x verify`'s `e2e` step printed `✓ e2e 46ms` and `"skipped": false` over a suite
+// whose only test skipped itself, because `bun test` exits 0 on a skip and the exit code was the
+// only thing the step's `ok` was read from. A gate that cannot tell a lane that ran from a lane
+// that did not is the one failure the step table exists to prevent, and it is the SAME question
+// the floor already answers one layer up (`X_VERIFY_SUITE_VANISHED`): the counts, not the code.
+describe('a suite that executed nothing is a skip, never a pass', () => {
+  // Bun's own summary for the scaffold's one e2e test on a box with no browser driver, verbatim.
+  // The reason lives in the test's NAME, which bun does not print — so the counts are the only
+  // channel, and this is exactly the output the gate had to read to be wrong about it.
+  const ALL_SKIPPED = ' 0 pass\n 1 skip\n 0 fail\nRan 1 test across 1 file. [46.00ms]';
+  const ONE_RAN = ' 1 pass\n 0 skip\n 0 fail\nRan 1 test across 1 file. [46.00ms]';
+
+  const suite =
+    (stdout: string): VerifyContext['runner'] =>
+    async (command) => ({
+      command,
+      code: 0,
+      ok: true,
+      stdout,
+      stderr: '',
+      durationMs: 46,
+    });
+
+  /** The real `e2e` step, over a root holding the file the scaffold writes for a new route. */
+  const e2eOver = async (stdout: string, root: string) => {
+    await Bun.write(join(root, 'apps/web/app/posts/page.e2e.test.ts'), '// e2e\n');
+    resetTestDiscovery();
+    const step = TEST_STEPS.find((candidate) => candidate.name === 'e2e');
+    if (step === undefined) return expect.unreachable('the e2e step is registered');
+    return runVerify([...STEPS.slice(0, 2), step], { root, runner: suite(stdout) });
+  };
+
+  const inTempRoot = async (run: (root: string) => Promise<void>): Promise<void> => {
+    const root = await mkdtemp(join(tmpdir(), 'x-verify-e2e-'));
+    try {
+      await run(root);
+    } finally {
+      resetTestDiscovery();
+      await rm(root, { recursive: true, force: true });
+    }
+  };
+
+  test('the step reports itself skipped, the way roadmap does', async () => {
+    await inTempRoot(async (root) => {
+      const result = await e2eOver(ALL_SKIPPED, root);
+      const e2e = result.steps?.find((step) => step.name === 'e2e');
+      expect(e2e?.skipped).toBe(true);
+      expect(dataOf(result.data).skipped).toEqual(['e2e']);
+    });
+  });
+
+  // Both renderers, because `--json` is what CI reads and the check mark is what a human reads,
+  // and the defect was visible in each.
+  test('neither renderer claims it passed', async () => {
+    await inTempRoot(async (root) => {
+      const result = await e2eOver(ALL_SKIPPED, root);
+      expect(result.summary).toContain('1 skipped: e2e');
+      expect(renderHuman(result)).toContain('- e2e');
+      expect(renderHuman(result)).not.toContain('✓ e2e');
+      const payload = JSON.parse(renderJson(result)) as {
+        readonly steps: readonly { readonly name: string; readonly skipped: boolean }[];
+      };
+      expect(payload.steps.find((step) => step.name === 'e2e')).toMatchObject({
+        skipped: true,
+        ok: true,
+        tests: { ran: 0, skipped: 1 },
+      });
+    });
+  });
+
+  // The other side of the enumeration: this may not turn a suite that ran into a skip.
+  test('one test that actually ran keeps the step a pass', async () => {
+    await inTempRoot(async (root) => {
+      const result = await e2eOver(ONE_RAN, root);
+      expect(result.steps?.find((step) => step.name === 'e2e')?.skipped).not.toBe(true);
+      expect(dataOf(result.data).skipped).toEqual([]);
+      expect(result.ok).toBe(true);
+    });
+  });
+});
diff --git a/packages/cli/src/verify-run.ts b/packages/cli/src/verify-run.ts
index c49dbb68..d55b34da 100644
--- a/packages/cli/src/verify-run.ts
+++ b/packages/cli/src/verify-run.ts
@@ -53,20 +53,29 @@ export async function runVerify(
         findings: [findingOf(error, step.name)],
       }),
     );
+    // A suite that executed nothing did not run, whatever its exit code says: `bun test` exits 0
+    // over an all-skipped file, so the counts are the only channel that can tell the two apart.
+    // ONE definition of "nothing ran", read twice, because the floor decides which of the two
+    // things it means — exactly as it already does for a step whose `applies` said no.
+    const tests = outcome.tests;
+    const nothingRan = tests !== undefined && tests.ran === 0;
+    const required = floorRequires(floor, step.name);
     // A step the floor requires whose suite executed nothing is the same vanished suite as a step
     // with no files at all — the run just had to finish before it could be seen. Appended to the
     // step's own findings so `data.failed`, the counts and every gate reading this table carry it.
-    const vanished =
-      floorRequires(floor, step.name) && outcome.tests !== undefined && outcome.tests.ran === 0
-        ? [skippedSuiteFinding(step.name, outcome.tests.skipped)]
-        : [];
+    const vanished = nothingRan && required ? [skippedSuiteFinding(step.name, tests.skipped)] : [];
     results.push({
       name: step.name,
       ok: outcome.ok && vanished.length === 0,
       durationMs: Math.round(performance.now() - started),
+      // Without a floor to require it, a suite that ran nothing is a SKIP and not a pass (#434):
+      // the `e2e` step printed `✓ e2e 46ms` over its one skipped test, which is the one thing a
+      // step table may never do — a reader cannot tell a lane that ran from a lane that did not.
+      skipped: nothingRan && !required,
       findings: [...outcome.findings, ...vanished],
       ...(outcome.output === undefined ? {} : { output: outcome.output }),
       ...(outcome.workers === undefined ? {} : { workers: outcome.workers }),
+      ...(tests === undefined ? {} : { tests }),
     });
   }
   const failedSteps = results.filter((step) => !step.ok).map((step) => step.name);
diff --git a/packages/testing/src/test-types.ts b/packages/testing/src/test-types.ts
index 0304186d..0a505ec3 100644
--- a/packages/testing/src/test-types.ts
+++ b/packages/testing/src/test-types.ts
@@ -96,13 +96,14 @@ let e2eDriver: ((name: string, body: E2eBody) => void) | undefined;
  * Register the browser-backed driver. Without one, `e2eTest` skips loudly — the skipped test's own
  * NAME carries the reason and the command that would build what it drives.
  *
- * What the gate then reports is a **pass over an all-skipped suite**, and that is stated here
- * rather than claimed away: this used to say "`x verify` reports the step as skipped rather than
- * green", which nothing implements. The step shells out to `bun test`, `bun test` exits 0 on a
- * skip, and an exit code is the only channel between the two — the driver is registered inside the
- * CHILD process, so the step cannot ask. Closing it means a channel the step can read, which is a
- * design decision and not a docstring. `As of 2026-08` there are zero registered drivers, so every
- * `e2eTest` in the tree is a skip; the framework's own `e2e` suites use plain `bun:test`.
+ * The gate's rule over that skip: a step whose suite reports `ran === 0` is SKIPPED, unless the
+ * repo's `x.verify.json` requires the step — then it is `X_VERIFY_SUITE_VANISHED`
+ * (`packages/cli/src/verify-run.ts`). Never a pass either way. This block used to say the opposite
+ * and argue for it — that an exit code was the only channel, so the gate could not tell — which
+ * `packages/cli/src/test-counts.ts` had already disproved by reading `bun test`'s own counts (#434).
+ *
+ * `As of 2026-08` there are zero registered drivers, so every `e2eTest` in the tree is a skip; the
+ * framework's own `e2e` suites use plain `bun:test`.
  */
 export function useE2eDriver(driver: (name: string, body: E2eBody) => void): void {
   e2eDriver = driver;