diff --git a/test/helpers/claude-pty-runner.ts b/test/helpers/claude-pty-runner.ts
index ef625a194..825b01bbf 100644
--- a/test/helpers/claude-pty-runner.ts
+++ b/test/helpers/claude-pty-runner.ts
@@ -26,6 +26,7 @@ import * as fs from 'fs';
 import * as os from 'os';
 import * as path from 'path';
 import { hermeticChildEnv, hermeticSkillsConfigDir, isHermeticEnabled } from './hermetic-env';
+import { createPlanCountFixture } from './plan-count-fixture';
 
 /** Strip ANSI escapes for pattern-matching against visible text. */
 export function stripAnsi(s: string): string {
@@ -1929,14 +1930,13 @@ export interface PlanSkillCountObservation {
  * AskUserQuestions until a terminal signal fires.
  *
  * Flow:
- *   1. Boot PTY in plan mode (8s grace + auto-trust dialog).
- *   2. Send `slashCommand` alone. Sleep ~3s.
- *   3. Send `followUpPrompt` as a chat message — this is the plan content
- *      the skill reviews. Slash commands with trailing args are rejected by
- *      Claude Code unless the skill defines them, so the plan goes as a
- *      follow-up message (the proven pattern at
- *      skill-e2e-plan-design-with-ui.test.ts:57-71).
- *   4. Poll loop:
+ *   1. Seed the complete fixture request in an isolated git repository's
+ *      PLAN.md and initial CLAUDE.md context, then boot the PTY in that cwd
+ *      (8s grace + auto-trust dialog). Skills remain registered in user scope.
+ *   2. Send `slashCommand` alone. The fixture is already in context; sending
+ *      it later can queue it behind the skill's first question while the
+ *      review incorrectly starts against the operator's live branch.
+ *   3. Poll loop:
  *      - Skip permission dialogs (auto-grant with `defaultPick`).
  *      - On a new numbered-option list, parse prompt + options, build
  *        fingerprint via `auqFingerprint`. Empty-prompt parses are skipped
@@ -1970,7 +1970,7 @@ export async function runPlanSkillCounting(opts: {
   skillName: string;
   /** Slash command to send alone, e.g. '/plan-ceo-review'. No trailing args. */
   slashCommand: string;
-  /** Plan content sent as a follow-up message ~3s after the slash command. */
+  /** Fixture request seeded in initial project context before the slash command. */
   followUpPrompt: string;
   /** Per-skill predicate: which answered AUQ is the last Step-0 question. */
   isLastStep0AUQ: Step0BoundaryPredicate;
@@ -1989,11 +1989,9 @@ export async function runPlanSkillCounting(opts: {
    * interview". Pressing the default 1 routes to "branch diff" (the wrong
    * review target for a seeded fixture). firstAUQPick lets the test pick
    * "Skip interview" or "describe inline" so the agent reviews the
-   * follow-up plan content the test sent, not the git diff.
+   * fixture plan content, not the git diff.
    */
   firstAUQPick?: (fp: AskUserQuestionFingerprint) => number;
-  /** Working directory. Default process.cwd() (repo cwd holds skill registry). */
-  cwd?: string;
   /** Total budget for skill to reach a terminal outcome. Default 1_500_000 (25 min). */
   timeoutMs?: number;
   /** Extra env merged into the spawned `claude` process. */
@@ -2005,14 +2003,21 @@ export async function runPlanSkillCounting(opts: {
   const defaultPick = opts.defaultPick ?? 1;
   const timeoutMs = opts.timeoutMs ?? 1_500_000;
 
-  const session = await launchClaudePty({
-    permissionMode: 'plan',
-    cwd: opts.cwd,
-    timeoutMs: timeoutMs + 60_000,
-    env: opts.env,
-    model: opts.model,
-    seedSkills: true,
-  });
+  const fixture = createPlanCountFixture(opts.followUpPrompt);
+  let session: ClaudePtySession;
+  try {
+    session = await launchClaudePty({
+      permissionMode: 'plan',
+      cwd: fixture.cwd,
+      timeoutMs: timeoutMs + 60_000,
+      env: opts.env,
+      model: opts.model,
+      seedSkills: true,
+    });
+  } catch (error) {
+    fixture.cleanup();
+    throw error;
+  }
 
   const fingerprints: AskUserQuestionFingerprint[] = [];
   const seen = new Set<string>();
@@ -2042,8 +2047,6 @@ export async function runPlanSkillCounting(opts: {
     await Bun.sleep(8000); // boot grace + auto-trust handler window
     const since = session.mark();
     session.send(`${opts.slashCommand}\r`);
-    await Bun.sleep(3000);
-    session.send(`${opts.followUpPrompt}\r`);
 
     const budgetStart = Date.now();
     while (Date.now() - budgetStart < timeoutMs) {
@@ -2170,7 +2173,11 @@ export async function runPlanSkillCounting(opts: {
       session.visibleSince(since),
     );
   } finally {
-    await session.close();
+    try {
+      await session.close();
+    } finally {
+      fixture.cleanup();
+    }
   }
 }
 
diff --git a/test/helpers/plan-count-fixture.ts b/test/helpers/plan-count-fixture.ts
new file mode 100644
index 000000000..7570a6e76
--- /dev/null
+++ b/test/helpers/plan-count-fixture.ts
@@ -0,0 +1,53 @@
+import * as fs from 'node:fs';
+import * as os from 'node:os';
+import * as path from 'node:path';
+import { spawnSync } from 'node:child_process';
+
+/**
+ * Count evals review a seeded plan, never the checkout that supplies skills.
+ * Put the complete request in Claude's initial project context before the
+ * bare slash command starts: a later message can remain queued behind the
+ * skill's first AskUserQuestion and leave it reviewing the live branch.
+ */
+export function createPlanCountFixture(prompt: string): {
+  cwd: string;
+  cleanup(): void;
+} {
+  const cwd = fs.mkdtempSync(path.join(os.tmpdir(), 'gstack-plan-count-'));
+  const cleanup = () => fs.rmSync(cwd, { recursive: true, force: true });
+  try {
+    fs.writeFileSync(path.join(cwd, 'PLAN.md'), prompt);
+    fs.writeFileSync(path.join(cwd, 'CLAUDE.md'), [
+      '# Plan review fixture',
+      '',
+      'This repository contains the plan under review. Use PLAN.md as the',
+      'current plan for the requested plan-review skill. The skill installation',
+      'supplies the workflow; its source checkout is not the review target.',
+      '',
+      'The complete user request is available from the start of this session:',
+      '',
+      prompt,
+      '',
+    ].join('\n'));
+
+    const git = (args: string[]) => {
+      const result = spawnSync('git', args, {
+        cwd,
+        encoding: 'utf8',
+        timeout: 10_000,
+      });
+      if (result.error || result.status !== 0) {
+        throw new Error(`Could not initialize plan-count fixture: ${result.error?.message ?? result.stderr}`);
+      }
+    };
+    git(['init', '-b', 'main']);
+    git(['add', 'PLAN.md', 'CLAUDE.md']);
+    git(['-c', 'user.name=Plan Count Fixture', '-c', 'user.email=plan-count@example.test',
+      '-c', 'commit.gpgsign=false', 'commit', '--no-verify', '-m', 'Seed review plan']);
+    git(['update-ref', 'refs/remotes/origin/main', 'HEAD']);
+    return { cwd, cleanup };
+  } catch (error) {
+    cleanup();
+    throw error;
+  }
+}
diff --git a/test/helpers/touchfiles-data.ts b/test/helpers/touchfiles-data.ts
index 3d609a6be..cf2dce00d 100644
--- a/test/helpers/touchfiles-data.ts
+++ b/test/helpers/touchfiles-data.ts
@@ -160,10 +160,10 @@ export const E2E_TOUCHFILES: Record<string, string[]> = {
   // Each test drives its skill end-to-end; touchfiles include preamble +
   // completion-status resolvers because they affect question cadence and
   // terminal output (the regression surface this test catches).
-  'plan-ceo-finding-count':      ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-ceo-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/skill-e2e-plan-ceo-finding-count.test.ts'],
-  'plan-eng-finding-count':      ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-eng-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/skill-e2e-plan-eng-finding-count.test.ts'],
-  'plan-design-finding-count':   ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-design-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/skill-e2e-plan-design-finding-count.test.ts'],
-  'plan-devex-finding-count':    ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-devex-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/skill-e2e-plan-devex-finding-count.test.ts'],
+  'plan-ceo-finding-count':      ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-ceo-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/skill-e2e-plan-ceo-finding-count.test.ts'],
+  'plan-eng-finding-count':      ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-eng-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/skill-e2e-plan-eng-finding-count.test.ts'],
+  'plan-design-finding-count':   ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-design-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/skill-e2e-plan-design-finding-count.test.ts'],
+  'plan-devex-finding-count':    ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-devex-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/skill-e2e-plan-devex-finding-count.test.ts'],
 
   // Gate-tier reviewCount-floor counterparts. Catch the May 2026 transcript
   // bug (model wrote a plan-mode plan and ExitPlanMode'd without firing any
@@ -180,8 +180,8 @@ export const E2E_TOUCHFILES: Record<string, string[]> = {
   // a model fires one AUQ then batches the rest into a "## Decisions to
   // confirm" plan write. runPlanSkillFloorCheck cannot detect that shape
   // (it exits on first AUQ); runPlanSkillCounting can.
-  'plan-eng-multi-finding-batching': ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-eng-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'scripts/resolvers/review.ts', 'test/helpers/claude-pty-runner.ts', 'test/fixtures/forcing-finding-seeds.ts', 'test/skill-e2e-plan-eng-multi-finding-batching.test.ts'],
-  'plan-ceo-split-overflow': ['plan-ceo-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'bin/gstack-question-preference', 'test/helpers/claude-pty-runner.ts', 'test/fixtures/forcing-finding-seeds.ts', 'test/skill-e2e-plan-ceo-split-overflow.test.ts'],
+  'plan-eng-multi-finding-batching': ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'plan-eng-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'scripts/resolvers/preamble/generate-completion-status.ts', 'scripts/resolvers/review.ts', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/fixtures/forcing-finding-seeds.ts', 'test/skill-e2e-plan-eng-multi-finding-batching.test.ts'],
+  'plan-ceo-split-overflow': ['plan-ceo-review/**', 'scripts/resolvers/preamble.ts', 'scripts/resolvers/preamble/generate-ask-user-format.ts', 'bin/gstack-question-preference', 'test/helpers/claude-pty-runner.ts', 'test/helpers/plan-count-fixture.ts', 'test/plan-count-fixture.test.ts', 'test/fixtures/forcing-finding-seeds.ts', 'test/skill-e2e-plan-ceo-split-overflow.test.ts'],
   'brain-privacy-gate':           ['bin/gstack-skill-start', 'bin/gstack-skill-end', 'scripts/resolvers/preamble/generate-brain-sync-block.ts', 'scripts/resolvers/preamble.ts', 'bin/gstack-brain-sync', 'bin/gstack-artifacts-init', 'bin/gstack-config', 'test/helpers/agent-sdk-runner.ts', 'test/skill-e2e-brain-privacy-gate.test.ts'],
 
   // /setup-gbrain Path 4 (Remote MCP) — happy + bad-token end-to-end via
diff --git a/test/plan-count-fixture.test.ts b/test/plan-count-fixture.test.ts
new file mode 100644
index 000000000..61e80199e
--- /dev/null
+++ b/test/plan-count-fixture.test.ts
@@ -0,0 +1,222 @@
+/** Free behavioral coverage for the isolated, preloaded plan-count workspace. */
+import { describe, expect, test } from 'bun:test';
+import { spawnSync } from 'node:child_process';
+import * as fs from 'node:fs';
+import * as os from 'node:os';
+import * as path from 'node:path';
+import { pathToFileURL } from 'node:url';
+import { createPlanCountFixture } from './helpers/plan-count-fixture';
+
+const ROOT = path.resolve(import.meta.dir, '..');
+const PROMPT = '# Seeded settings plan\n\nReview each issue separately.\n' +
+  'Literal text: "quotes" \'single quotes\' `touch never` $(touch never)\n';
+
+function gitRoot(cwd: string): string {
+  const result = spawnSync('git', ['rev-parse', '--show-toplevel'], { cwd, encoding: 'utf8', timeout: 10_000 });
+  expect(result.status, result.stderr).toBe(0);
+  return fs.realpathSync(result.stdout.trim());
+}
+
+describe('plan-count fixtures', () => {
+  test('preloads the literal plan and initial context in a separate git repository', () => {
+    const fixture = createPlanCountFixture(PROMPT);
+    try {
+      expect(fs.realpathSync(fixture.cwd)).not.toBe(fs.realpathSync(ROOT));
+      expect(gitRoot(fixture.cwd)).toBe(fs.realpathSync(fixture.cwd));
+      expect(fs.readFileSync(path.join(fixture.cwd, 'PLAN.md'), 'utf8')).toBe(PROMPT);
+      const context = fs.readFileSync(path.join(fixture.cwd, 'CLAUDE.md'), 'utf8');
+      expect(context).toContain('PLAN.md');
+      expect(context).toContain(PROMPT);
+      expect(fs.existsSync(path.join(fixture.cwd, 'DESIGN.md'))).toBe(false);
+      expect(fs.existsSync(path.join(fixture.cwd, 'TODOS.md'))).toBe(false);
+      expect(fs.existsSync(path.join(fixture.cwd, 'never'))).toBe(false);
+    } finally {
+      fixture.cleanup();
+    }
+    expect(fs.existsSync(fixture.cwd)).toBe(false);
+  });
+
+  test('concurrent fixtures have independent content and cleanup owns only its directory', () => {
+    const first = createPlanCountFixture('first plan');
+    const second = createPlanCountFixture('second plan');
+    try {
+      expect(first.cwd).not.toBe(second.cwd);
+      first.cleanup();
+      first.cleanup();
+      expect(fs.existsSync(first.cwd)).toBe(false);
+      expect(fs.readFileSync(path.join(second.cwd, 'PLAN.md'), 'utf8')).toBe('second plan');
+      expect(gitRoot(second.cwd)).toBe(fs.realpathSync(second.cwd));
+    } finally {
+      first.cleanup();
+      second.cleanup();
+    }
+    expect(fs.existsSync(second.cwd)).toBe(false);
+  });
+
+  test.skipIf(process.platform === 'win32')('cleans the fixture when the PTY executable cannot launch', () => {
+    const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'plan-count-launch-failure-'));
+    const fixtureTmp = path.join(dir, 'fixtures');
+    const workerPath = path.join(dir, 'worker.ts');
+    const brokenCli = path.join(dir, 'not-executable');
+    fs.mkdirSync(fixtureTmp);
+    fs.writeFileSync(brokenCli, 'This file deliberately has no executable permission.', { mode: 0o644 });
+    const runnerUrl = pathToFileURL(path.join(ROOT, 'test/helpers/claude-pty-runner.ts')).href;
+    fs.writeFileSync(workerPath, `
+import { runPlanSkillCounting } from ${JSON.stringify(runnerUrl)};
+try {
+  await runPlanSkillCounting({
+    skillName: 'plan-design-review', slashCommand: '/plan-design-review',
+    followUpPrompt: 'Review this launch-failure fixture.',
+    isLastStep0AUQ: () => false, reviewCountCeiling: 8,
+  });
+  process.exitCode = 1;
+} catch (error) {
+  process.stdout.write('launch failed as expected');
+}
+`);
+    try {
+      const result = spawnSync(process.execPath, [workerPath], {
+        cwd: ROOT,
+        env: { ...process.env, BROWSE_TERMINAL_BINARY: brokenCli, EVALS_HERMETIC: '1', TMPDIR: fixtureTmp },
+        encoding: 'utf8',
+        timeout: 10_000,
+      });
+      expect(result.status, result.stderr).toBe(0);
+      expect(result.stdout).toBe('launch failed as expected');
+      expect(fs.readdirSync(fixtureTmp)).toEqual([]);
+    } finally {
+      fs.rmSync(dir, { recursive: true, force: true });
+    }
+  }, 12_000);
+
+  // Bun's real PTY spawn and executable shebangs are POSIX-only. The direct
+  // fixture tests above still exercise repository setup/cleanup on Windows.
+  test.skipIf(process.platform === 'win32')(
+    'real PTY children see preloaded isolated plans, shipped skills, and only the bare slash command',
+    async () => {
+      const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'plan-count-subprocess-'));
+      const fakePath = path.join(dir, 'fake-claude');
+      const workerPath = path.join(dir, 'worker.ts');
+      const resultPath = path.join(dir, 'result.json');
+      const cases = [
+        { name: 'design', skillName: 'plan-design-review', prompt: PROMPT, mode: 'complete' },
+        { name: 'ceo', skillName: 'plan-ceo-review', prompt: '# Independent CEO plan\nUnique product context.', mode: 'complete' },
+        { name: 'exited', skillName: 'plan-eng-review', prompt: '# Early-exit plan\nStill clean up.', mode: 'exit' },
+      ].map((item) => ({ ...item, record: path.join(dir, `${item.name}.jsonl`) }));
+
+      // The fake snapshots its environment BEFORE installing the first stdin
+      // handler. It never dispatches a model or invokes the real Claude CLI.
+      fs.writeFileSync(fakePath, `#!${process.execPath}\n` + String.raw`
+import * as fs from 'node:fs';
+import * as path from 'node:path';
+import { spawnSync } from 'node:child_process';
+const record = (event) => fs.appendFileSync(process.env.FIXTURE_RECORD, JSON.stringify(event) + '\n');
+const skillDir = path.join(process.env.CLAUDE_CONFIG_DIR, 'skills', process.env.FIXTURE_SKILL);
+const planPath = path.join(process.cwd(), 'PLAN.md');
+const contextPath = path.join(process.cwd(), 'CLAUDE.md');
+record({
+  type: 'startup', pid: process.pid, cwd: process.cwd(), argv: process.argv.slice(2),
+  gitRoot: spawnSync('git', ['rev-parse', '--show-toplevel'], { encoding: 'utf8', timeout: 10_000 }).stdout.trim(),
+  plan: fs.existsSync(planPath) ? fs.readFileSync(planPath, 'utf8') : null,
+  context: fs.existsSync(contextPath) ? fs.readFileSync(contextPath, 'utf8') : null,
+  inheritedDesign: fs.existsSync('DESIGN.md'), inheritedTodos: fs.existsSync('TODOS.md'),
+  skill: fs.existsSync(path.join(skillDir, 'SKILL.md')) ? fs.readFileSync(path.join(skillDir, 'SKILL.md'), 'utf8') : null,
+  sections: fs.existsSync(path.join(skillDir, 'sections', 'review-sections.md'))
+    ? fs.readFileSync(path.join(skillDir, 'sections', 'review-sections.md'), 'utf8') : null,
+});
+process.stdin.setRawMode?.(true);
+let firstInput = true;
+let completion;
+process.stdin.on('data', (data) => {
+  record({ type: 'input', data: data.toString('utf8') });
+  if (!firstInput) return;
+  firstInput = false;
+  if (process.env.FIXTURE_MODE === 'exit') process.exit(7);
+  // Longer than the old 3 s delayed fixture send: record that regression
+  // even if the helper would otherwise return on our completion marker.
+  completion = setTimeout(() => process.stdout.write('\nGSTACK REVIEW REPORT\n'), 4100);
+});
+process.on('SIGINT', () => {
+  clearTimeout(completion);
+  record({ type: 'closed', at: Date.now() });
+  process.exit(0);
+});
+process.stdin.resume();
+`);
+      fs.chmodSync(fakePath, 0o755);
+      const runnerUrl = pathToFileURL(path.join(ROOT, 'test/helpers/claude-pty-runner.ts')).href;
+      fs.writeFileSync(workerPath, `
+import { runPlanSkillCounting } from ${JSON.stringify(runnerUrl)};
+const cases = ${JSON.stringify(cases)};
+const results = await Promise.all(cases.map(async (item) => ({
+  name: item.name,
+  observation: await runPlanSkillCounting({
+    skillName: item.skillName,
+    slashCommand: '/' + item.skillName,
+    followUpPrompt: item.prompt,
+    isLastStep0AUQ: () => false,
+    reviewCountCeiling: 8,
+    timeoutMs: 15000,
+    env: { FIXTURE_RECORD: item.record, FIXTURE_SKILL: item.skillName, FIXTURE_MODE: item.mode },
+  }),
+})));
+await Bun.write(${JSON.stringify(resultPath)}, JSON.stringify(results));
+`);
+
+      const startedAt = Date.now();
+      const child = Bun.spawn([process.execPath, workerPath], {
+        cwd: ROOT,
+        env: { ...process.env, BROWSE_TERMINAL_BINARY: fakePath, EVALS_HERMETIC: '1' },
+        stdout: 'pipe',
+        stderr: 'pipe',
+      });
+      const killer = setTimeout(() => child.kill('SIGKILL'), 35_000);
+      try {
+        const [exitCode, stdout, stderr] = await Promise.all([
+          child.exited,
+          new Response(child.stdout).text(),
+          new Response(child.stderr).text(),
+        ]);
+        expect(exitCode, stdout + stderr).toBe(0);
+        expect(Date.now() - startedAt).toBeLessThan(35_000);
+        const results = JSON.parse(fs.readFileSync(resultPath, 'utf8'));
+        const cwds = new Set<string>();
+        for (const item of cases) {
+          const events = fs.readFileSync(item.record, 'utf8').trim().split('\n').map((line) => JSON.parse(line));
+          const startup = events[0];
+          expect(startup.type).toBe('startup');
+          expect(startup.cwd).not.toBe(ROOT);
+          expect(startup.gitRoot).toBe(startup.cwd);
+          expect(startup.plan).toBe(item.prompt);
+          expect(startup.context).toContain('PLAN.md');
+          expect(startup.context).toContain(item.prompt);
+          expect(startup.argv[startup.argv.indexOf('--permission-mode') + 1]).toBe('plan');
+          expect(startup.inheritedDesign).toBe(false);
+          expect(startup.inheritedTodos).toBe(false);
+          expect(startup.skill).toContain(`name: ${item.skillName}`);
+          expect(startup.sections).toBe(fs.readFileSync(path.join(ROOT, item.skillName, 'sections/review-sections.md'), 'utf8'));
+          expect(events.filter((event) => event.type === 'input').map((event) => event.data).join(''))
+            .toBe(`/${item.skillName}\r`);
+          expect(fs.existsSync(startup.cwd)).toBe(false);
+          expect(() => process.kill(startup.pid, 0)).toThrow();
+          const result = results.find((result) => result.name === item.name);
+          expect(result.observation.outcome).toBe(item.mode === 'exit' ? 'exited' : 'completion_summary');
+          if (item.mode === 'complete') expect(events.at(-1).type).toBe('closed');
+          cwds.add(startup.cwd);
+        }
+        expect(cwds.size).toBe(cases.length);
+      } finally {
+        clearTimeout(killer);
+        child.kill('SIGKILL');
+        // A failed worker assertion must not leave its PTY fakes running.
+        for (const item of cases) {
+          if (!fs.existsSync(item.record)) continue;
+          const startup = JSON.parse(fs.readFileSync(item.record, 'utf8').split('\n')[0]);
+          try { process.kill(startup.pid, 'SIGKILL'); } catch { /* already reaped */ }
+        }
+        fs.rmSync(dir, { recursive: true, force: true });
+      }
+    },
+    40_000,
+  );
+});
diff --git a/test/skill-e2e-autoplan-chain.test.ts b/test/skill-e2e-autoplan-chain.test.ts
index 72cf0513e..9d9e669c7 100644
--- a/test/skill-e2e-autoplan-chain.test.ts
+++ b/test/skill-e2e-autoplan-chain.test.ts
@@ -32,6 +32,7 @@ import { spawnSync } from 'child_process';
 import * as fs from 'fs';
 import * as path from 'path';
 import * as os from 'os';
+import { stripVTControlCharacters } from 'node:util';
 import {
   launchClaudePty,
   isPlanReadyVisible,
@@ -49,6 +50,12 @@ interface PhaseHit {
   ts: number;
 }
 
+function diagnosticTail(text: string): string {
+  return stripVTControlCharacters(text)
+    .replace(/[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]/g, '')
+    .slice(-3000);
+}
+
 describeE2E('/autoplan chain ordering (periodic)', () => {
   test(
     'phases run sequentially: Phase 1 (CEO) before Phase 3 (Eng), Phase 2 (Design) between when present',
@@ -79,6 +86,8 @@ describeE2E('/autoplan chain ordering (periodic)', () => {
         const hits: PhaseHit[] = [];
         let outcome: 'chain_complete' | 'plan_ready' | 'timeout' | 'exited' = 'timeout';
         let evidence = '';
+        let fullSessionEvidence = '';
+        let exitCode: number | null = null;
 
         try {
           await Bun.sleep(8000);
@@ -144,13 +153,18 @@ describeE2E('/autoplan chain ordering (periodic)', () => {
             }
           }
         } finally {
+          // Preserve boot failures omitted by mark(), and observe the real exit
+          // status before close() deliberately terminates a live session.
+          exitCode = session.exitCode();
+          fullSessionEvidence = diagnosticTail(session.visibleText());
           await session.close();
         }
 
         if (outcome === 'exited' || outcome === 'timeout') {
           throw new Error(
-            `autoplan chain test FAILED: outcome=${outcome}, hits=${JSON.stringify(hits)}\n` +
-              `--- evidence (last 3KB) ---\n${evidence}`,
+            `autoplan chain test FAILED: outcome=${outcome}, exitCode=${exitCode}, hits=${JSON.stringify(hits)}\n` +
+              `--- post-command evidence (last 3KB) ---\n${diagnosticTail(evidence)}\n` +
+              `--- full-session visible tail, including startup (last 3KB) ---\n${fullSessionEvidence}`,
           );
         }
 
diff --git a/test/skill-e2e-plan-ceo-finding-count.test.ts b/test/skill-e2e-plan-ceo-finding-count.test.ts
index 4b8038b02..6ed054bc2 100644
--- a/test/skill-e2e-plan-ceo-finding-count.test.ts
+++ b/test/skill-e2e-plan-ceo-finding-count.test.ts
@@ -38,8 +38,8 @@ import {
  * The default pick (1) routes to "branch diff vs main" — the wrong target
  * for our seeded fixture (the agent would review the gstack PR itself,
  * recursively). Picking "Skip interview and plan immediately" bypasses
- * Step 0 and routes the agent to review the chat context (where our
- * follow-up plan was pasted).
+ * Step 0 and routes the agent to review the fixture request, which is
+ * already present in its initial project context.
  */
 function pickSkipInterview(fp: AskUserQuestionFingerprint): number {
   const skipOpt = fp.options.find((o) =>
@@ -122,9 +122,6 @@ describeE2E('/plan-ceo-review per-finding AskUserQuestion count (periodic)', ()
           isLastStep0AUQ: ceoStep0Boundary,
           reviewCountCeiling: CEILING_DISTINCT + 1, // hard cap above assertion ceiling
           firstAUQPick: pickSkipInterview, // bypass scope-selection, route to review
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000, // 25 min
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
@@ -211,9 +208,6 @@ describeE2E('/plan-ceo-review per-finding AskUserQuestion count (periodic)', ()
           followUpPrompt: planCeo2PairedFindings(planPath),
           isLastStep0AUQ: ceoStep0Boundary,
           reviewCountCeiling: CEILING_PAIRED + 1,
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000,
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
diff --git a/test/skill-e2e-plan-ceo-split-overflow.test.ts b/test/skill-e2e-plan-ceo-split-overflow.test.ts
index 9f9060d68..94bc9abc4 100644
--- a/test/skill-e2e-plan-ceo-split-overflow.test.ts
+++ b/test/skill-e2e-plan-ceo-split-overflow.test.ts
@@ -73,9 +73,6 @@ describeE2E('/plan-ceo-review split-overflow regression (periodic)', () => {
           followUpPrompt,
           isLastStep0AUQ: ceoStep0Boundary,
           reviewCountCeiling: N + 3, // hard cap above floor + tolerance
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000, // 25 min
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
diff --git a/test/skill-e2e-plan-design-finding-count.test.ts b/test/skill-e2e-plan-design-finding-count.test.ts
index 6f2398d99..073b68470 100644
--- a/test/skill-e2e-plan-design-finding-count.test.ts
+++ b/test/skill-e2e-plan-design-finding-count.test.ts
@@ -69,9 +69,6 @@ describeE2E('/plan-design-review per-finding AskUserQuestion count (periodic)',
           followUpPrompt: planDesign5Findings(planPath),
           isLastStep0AUQ: designStep0Boundary,
           reviewCountCeiling: CEILING + 1,
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000,
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
@@ -94,11 +91,16 @@ describeE2E('/plan-design-review per-finding AskUserQuestion count (periodic)',
         if (obs.reviewCount < FLOOR) {
           throw new Error(
             `BAND FAIL (below floor): reviewCount=${obs.reviewCount} < FLOOR=${FLOOR}.\n` +
-              `Likely batching regression. Review-phase fingerprints:\n` +
+              `outcome=${obs.outcome} step0=${obs.step0Count} elapsed=${obs.elapsedMs}ms\n` +
+              `summary: ${obs.summary}\n` +
+              `All captured fingerprints (including Step 0):\n` +
               obs.fingerprints
-                .filter((f) => !f.preReview)
-                .map((f) => `  - "${f.promptSnippet.slice(0, 80)}"`)
-                .join('\n'),
+                .map(
+                  (f) =>
+                    `  - preReview=${f.preReview} sig=${f.signature.slice(0, 12)} prompt="${f.promptSnippet}"`,
+                )
+                .join('\n') +
+              `\n--- evidence (last 3KB) ---\n${obs.evidence}`,
           );
         }
         if (obs.reviewCount > CEILING) {
diff --git a/test/skill-e2e-plan-devex-finding-count.test.ts b/test/skill-e2e-plan-devex-finding-count.test.ts
index e80bdac5d..1e4edb068 100644
--- a/test/skill-e2e-plan-devex-finding-count.test.ts
+++ b/test/skill-e2e-plan-devex-finding-count.test.ts
@@ -69,9 +69,6 @@ describeE2E('/plan-devex-review per-finding AskUserQuestion count (periodic)', (
           followUpPrompt: planDevex5Findings(planPath),
           isLastStep0AUQ: devexStep0Boundary,
           reviewCountCeiling: CEILING + 1,
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000,
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
diff --git a/test/skill-e2e-plan-eng-finding-count.test.ts b/test/skill-e2e-plan-eng-finding-count.test.ts
index f51018c77..566776dd9 100644
--- a/test/skill-e2e-plan-eng-finding-count.test.ts
+++ b/test/skill-e2e-plan-eng-finding-count.test.ts
@@ -68,9 +68,6 @@ describeE2E('/plan-eng-review per-finding AskUserQuestion count (periodic)', ()
           followUpPrompt: planEng5Findings(planPath),
           isLastStep0AUQ: engStep0Boundary,
           reviewCountCeiling: CEILING + 1,
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000,
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });
diff --git a/test/skill-e2e-plan-eng-multi-finding-batching.test.ts b/test/skill-e2e-plan-eng-multi-finding-batching.test.ts
index 89d2b594d..d6c423d36 100644
--- a/test/skill-e2e-plan-eng-multi-finding-batching.test.ts
+++ b/test/skill-e2e-plan-eng-multi-finding-batching.test.ts
@@ -65,9 +65,6 @@ describeE2E('/plan-eng-review multi-finding batching regression (periodic)', ()
           followUpPrompt,
           isLastStep0AUQ: engStep0Boundary,
           reviewCountCeiling: N + 3, // hard cap above floor + tolerance
-          // LIVE-REPO CWD: PTY session needs the repo cwd — gstack skill
-          // registry + hermetic pre-trusted dir (hermetic-env trustedDirs).
-          cwd: process.cwd(),
           timeoutMs: 1_500_000, // 25 min
           env: { QUESTION_TUNING: 'false', EXPLAIN_LEVEL: 'default' },
         });