Files
mokosh/.planning/phases/03-spec-10-smoke-verification-dom-event-log-verification/03-PATTERNS.md

40 KiB
Raw Permalink Blame History

Phase 3: SPEC §10 smoke verification + DOM/event-log verification - Pattern Map

Mapped: 2026-05-20 Files analyzed: 5 anticipated (4 modifications + 1 new VERIFICATION.md) + optional 6th (Page.metrics scaffolding) Analogs found: 5 / 5 (4 exact, 1 role-match; +1 partial for optional Page.metrics) Phase character: Verification-heavy; Approach B harness extension (Plan 02-04 is direct precedent for 4 of 5 files).

File Classification

Anticipated File (Plans 03-01..05) Role Data Flow Closest Analog Match Quality
tests/uat/extension-page-harness.ts (page-side assertA29..A3X) test (page-side assertion host) request-response (chrome.runtime.sendMessage round-trip) + event-driven (chrome.downloads.onCreated listener for A24 echo, if needed by Plan 03-02) tests/uat/extension-page-harness.ts:2851-3413 (assertA24..assertA28 block; Plan 02-04) EXACT (same file; append-only)
tests/uat/lib/harness-page-driver.ts (host-side driveA29..A3X + JSZip reads) test (host-side driver + zip inspection) file-I/O (downloadsDir polling + JSZip-parse) + request-response (page.evaluate wrapper) tests/uat/lib/harness-page-driver.ts:1202-1849 (driveA24..driveA28 + findLatestZip + A28_EXPECTED_PATHS; Plan 02-04) EXACT (same file; append-only)
tests/uat/harness.test.ts (orchestrator wiring: imports + wrapped drivers + drivers array push + banner) test (orchestrator) request-response (sequential driver dispatch + bail-on-first-failure) tests/uat/harness.test.ts:93-99 + 316-335 + 406-430 (driveA24-A28 imports + wrappers + push; Plan 02-04) EXACT (same file; append-only)
tests/uat/extension-page-harness.html (probe HTML append) test (DOM scaffold) static markup tests/uat/extension-page-harness.html:17-22 (existing <h1> + <pre id="status">; Plan 01-12 Wave 6 tokens.css preserved) EXACT (same file; append-only)
.planning/phases/03-.../03-VERIFICATION.md (NEW; aggregator) docs (verification aggregator with frontmatter) docs synthesis .planning/phases/02-stabilize-export-pipeline/02-VERIFICATION.md (T5 override pattern; 5/5 must-haves; Plan 02-04 closure) PLUS .planning/phases/01-stabilize-video-pipeline/01-VERIFICATION.md (per-requirement scorecard; cross-cutting gates table; operator empirical acks; deferred items section) EXACT (T5 override) + ROLE-MATCH (scorecard structure)
tests/uat/lib/ramMetrics.ts OR inline in harness-page-driver.ts (OPTIONAL — D-P3-04 best-effort puppeteer.Page.metrics() scaffolding) test (host-side utility) request-response (CDP Performance.getMetrics via Page.metrics()) NONE (research-derived; no existing analog) — closest scaffolding shape lives in tests/uat/lib/harness-page-driver.ts:1255-1334 (driveA25 host-side latency measurement + diagnostic copy pattern) NEW PATTERN (cite RESEARCH §"Code Examples" A3X)

Pattern Assignments

tests/uat/extension-page-harness.ts (page-side assertA29..A3X for Plans 03-01..04)

Analog: tests/uat/extension-page-harness.ts:2851-3413 (Plan 02-04 assertA24-A28)

Existing imports / module context — already present at top of file; no new imports expected:

  • The harness uses ambient chrome.* types (chrome.downloads.DownloadItem, chrome.tabs.create, etc.); production surfaces only.
  • Helper utilities setupFreshRecording, sendMessageWithTimeout, diag (file-local) — all reusable for Plan 03-01..04.

Module constants block — Plan 02-04 precedent for per-assertion tuning constants (lines 2977-2982):

/** SAVE_ARCHIVE dispatch timeout for A25 — matches A24's 15s. */
const A25_SAVE_ARCHIVE_TIMEOUT_MS = 15_000;
/** Pre-SAVE segment-settle window (10s rotation + 1s slack). */
const A25_SEGMENT_SETTLE_MS = 11_000;
/** Hard latency ceiling per SPEC §10 #6 + CON-archive-export-latency. */
const A25_LATENCY_CEILING_MS = 5_000;

Phase 3 plans should follow the same const A<NN>_<PURPOSE>_MS = NNNN_NNN; form (A29_SEGMENT_SETTLE_MS, A30_TRIGGER_PAUSE_MS, A31_PASSWORD_SENTINEL, etc.).

Page-side stub pattern (for host-side-only work)assertA26 lines 3164-3173 + assertA28 lines 3307-3316:

/**
 * A26 — D-P2-02 + D-P2-03 empirical (page-side stub).
 *
 * Returns the assertion name + a sentinel diagnostic. All real work
 * happens host-side in driveA26 (JSZip-parse the latest zip + assert
 * meta.json shape). The page-side stub exists purely so the orchestrator's
 * single-method-per-assertion contract (window.__mokoshHarness.assertA26)
 * is uniform across all 29 assertions.
 *
 * Chaining: A26 reads the zip produced by A25 (host-side driver picks
 * the most-recently-modified zip in downloadsDir). No new SAVE dispatch.
 *
 * @returns Stub AssertionResult with empty checks; driveA26 fills them.
 */
async function assertA26(): Promise<AssertionResult> {
  return {
    passed: true,
    name: 'A26 — meta.json 8-field shape (D-P2-02 + D-P2-03)',
    checks: [],
    diagnostics: [
      'assertA26 page-side stub; host-side driveA26 inspects latest zip + asserts meta.json shape',
    ],
  };
}

Use for: Plan 03-01 (rrweb session.json inspection is host-side JSZip), Plan 03-02 (events.json grep is host-side), Plan 03-03 (sentinel grep is host-side). Page-side stubs for A29 / A30 / A31; host-side drivers do the work.

Page-side orchestration pattern (when chrome. APIs are required pre-SAVE)* — assertA24 lines 2851-2962 (chrome.downloads.onCreated cross-realm capture) and assertA27 lines 3202-3292 (chrome.tabs.create + activate + cleanup):

async function assertA24(): Promise<AssertionResult> {
  const result: AssertionResult = {
    passed: false,
    name: 'A24 — D-P2-01 empirical: chrome.downloads.download receives blob: URL (closes P0-6)',
    checks: [],
    diagnostics: [],
  };

  // Capture stored in closure so the onCreated listener can populate it
  // across the async SAVE_ARCHIVE dispatch + post-ack poll.
  let capturedUrl: string | null = null;
  let listenerInstalled = false;
  const onCreatedListener = (item: chrome.downloads.DownloadItem): void => {
    if (capturedUrl === null) {
      capturedUrl = item.url;
    }
  };

  try {
    diag(result, 'Step 1: setupFreshRecording (A24 owns its recording — onCreated listener installed pre-SAVE)');
    const setupResp = await setupFreshRecording();
    if (!setupResp.ok) {
      throw new Error(`setupFreshRecording failed: ${setupResp.error ?? '(no error)'}`);
    }
    // ... Step 2: settle, Step 3: install listener, Step 4: dispatch SAVE_ARCHIVE, Step 5: poll for capturedUrl
  } catch (err) {
    result.error = err instanceof Error ? err.message : String(err);
    diag(result, `THREW: ${result.error}`);
  } finally {
    // T-02-04-01 mitigation: always remove the listener (idempotent; no-op if never added).
    if (listenerInstalled) {
      try { chrome.downloads.onCreated.removeListener(onCreatedListener); }
      catch (rmErr) { diag(result, `(listener cleanup ignored: ${String(rmErr)})`); }
    }
  }

  return result;
}

Use for: Plan 03-02 if event-log triggers (page.click / page.type / dispatchEvent) require pre-SAVE setup beyond what RESEARCH §"Code Examples" pattern 2 describes; Plan 03-03 password-sentinel typing pre-SAVE.

Page-side extended-result shape (when host needs page data beyond AssertionResult)A25Result lines 2997-3002 + A27Result lines 3145-3148:

interface A25Result extends AssertionResult {
  t0: number;        // performance.now() at SAVE dispatch
  tAck: number;      // performance.now() at SAVE ack
  t0Wall: number;    // Date.now() at SAVE dispatch (cross-realm bracket anchor)
  ackSuccess: boolean;
}

interface A27Result extends AssertionResult {
  tabAUrl: string;
  tabBUrl: string;
}

Use for: Plan 03-04 OPTIONAL Page.metrics scaffolding — A3XResult extends AssertionResult { jsHeapUsedBytes: number; jsHeapTotalBytes: number; } if Plan 03-04 ships scaffolding (D-P3-04 "if practical"; see RESEARCH §"Code Example A3X").

__mokoshHarness registration pattern — lines 3332-3372 (TypeScript ambient interface) + lines 3374-3404 (object literal):

declare global {
  interface Window {
    __mokoshHarness: {
      assertA1: () => Promise<AssertionResult>;
      // ... (24 baseline assertions A1..A23)
      assertA24: () => Promise<AssertionResult>;
      assertA25: () => Promise<A25Result>;
      assertA26: () => Promise<AssertionResult>;
      assertA27: () => Promise<A27Result>;
      assertA28: () => Promise<AssertionResult>;
      getManifestVersion: () => Promise<string>;
    };
  }
}

window.__mokoshHarness = {
  assertA1, assertA2, /* ... */ assertA27, assertA28, getManifestVersion,
};

Use for: Plans 03-01..04 each append their own assertA<NN> to BOTH the declare global interface AND the window.__mokoshHarness = {...} object literal. Lockstep is critical; missing one side breaks the orchestrator's await harness.assertA<NN>() call.


tests/uat/lib/harness-page-driver.ts (host-side driveA29..A3X + JSZip)

Analog: tests/uat/lib/harness-page-driver.ts:1202-1849 (Plan 02-04 driveA24-A28)

Existing imports (already present in file) — lines 36-45:

import { spawnSync } from 'node:child_process';
import { existsSync, mkdtempSync, readFileSync, readdirSync, statSync, unlinkSync } from 'node:fs';
import { tmpdir } from 'node:os';
import { join, resolve as resolvePath } from 'node:path';

import JSZip from 'jszip';
import type { Page } from 'puppeteer';

import type { AssertionRecord, CheckRecord } from './assertions';
import { assertArchiveShape, extractEntryToFile } from './zip';

Plan 03-01 only needs to ADD a single import for the rrweb EventType enum:

import { EventType } from '@rrweb/types';   // ← NEW for Plan 03-01 (verified at node_modules/@rrweb/types/dist/index.d.ts:186-194)

Plan 03-02 / 03-03 may add an import for UserEvent from src/shared/types.ts:

import type { UserEvent } from '../../../src/shared/types';   // ← NEW for Plan 03-02 events.json grep

Standard page.evaluate wrapper for page-orchestrated assertions — driveA24 lines 1202-1209:

export async function driveA24(page: Page): Promise<AssertionRecord> {
  return await page.evaluate(async () => {
    // eslint-disable-next-line @typescript-eslint/no-explicit-any -- evaluate runs in browser context where Window types are loose.
    const harness = (window as any).__mokoshHarness;
    const r: AssertionRecord = await harness.assertA24();
    return r;
  }) as AssertionRecord;
}

Use for: Plan 03-02 driveA30 (if Plan 03-02 chooses page-side orchestration for event triggers — but RESEARCH recommends Puppeteer-side page.click / page.type via the page, not chrome.runtime.sendMessage round-trip; planner picks).

Two-phase pattern (page-side stub + host-side JSZip read) — driveA26 lines 1421-1567 (canonical 6-check meta.json shape gate; chains off A25):

export async function driveA26(
  page: Page,
  downloadsDir: string,
): Promise<AssertionRecord> {
  // Phase 1 — page-side stub (uniform orchestrator shape).
  const pageResult = await page.evaluate(async () => {
    // eslint-disable-next-line @typescript-eslint/no-explicit-any -- evaluate runs in browser context where Window types are loose.
    const harness = (window as any).__mokoshHarness;
    const r: AssertionRecord = await harness.assertA26();
    return r;
  }) as AssertionRecord;

  const mergedChecks: CheckRecord[] = pageResult.checks.slice();
  const mergedDiagnostics: string[] = pageResult.diagnostics.slice();

  // Phase 2 — locate A25's zip.
  const zipPath = findLatestZip(downloadsDir);
  if (zipPath === null) {
    mergedChecks.push({
      name: 'A26.0: at least one zip present in downloadsDir (chain-off A25)',
      expected: '>=1 zip',
      actual: 'no zip in downloadsDir',
      passed: false,
    });
    return { passed: false, name: pageResult.name, checks: mergedChecks,
             diagnostics: mergedDiagnostics, error: pageResult.error };
  }
  mergedDiagnostics.push(`A26 zipPath=${zipPath}`);

  // Phase 3 — load + inspect target file inside zip.
  const zipBytes = readFileSync(zipPath);
  const zip = await JSZip.loadAsync(zipBytes);
  const metaFile = zip.file('meta.json');
  // ... A26.1..A26.6 structural checks
  const mergedPassed = mergedChecks.every((c) => c.passed);
  return { passed: mergedPassed, name: pageResult.name, checks: mergedChecks,
           diagnostics: mergedDiagnostics, error: pageResult.error };
}

Use for: All four Plans 03-01..04 — same 3-phase shape. Plan 03-01 reads rrweb/session.json, Plan 03-02 reads logs/events.json, Plan 03-03 reads logs/events.json for sentinel grep.

findLatestZip chain-off helper — lines 1395-1406 (must be present; if Plan 03-* runs in a fresh wave the helper is already shipped by Plan 02-04):

function findLatestZip(downloadsDir: string): string | null {
  const candidates = readdirSync(downloadsDir).filter(isZipFilename);
  if (candidates.length === 0) {
    return null;
  }
  const withMtimes = candidates.map((name) => ({
    name,
    mtimeMs: statSync(resolvePath(downloadsDir, name)).mtimeMs,
  }));
  withMtimes.sort((a, b) => b.mtimeMs - a.mtimeMs);
  return resolvePath(downloadsDir, withMtimes[0].name);
}

Use for: Every Plan 03-01..04 driver that reads from downloadsDir. Already present in file (no re-implementation).

EventType enum grep pattern (Plan 03-01) — derived from RESEARCH §"Code Examples A29" (verified against @rrweb/types/dist/index.d.ts:186-194):

// Plan 03-01 driveA29 — structural rrweb event-shape grep (RESEARCH-cited pattern; no codebase analog)
import { EventType } from '@rrweb/types';   // FullSnapshot=2, IncrementalSnapshot=3, Meta=4

const rrwebRaw = await zip.file('rrweb/session.json')!.async('text');
const events: Array<{ type: number; timestamp: number }> = JSON.parse(rrwebRaw);

mergedChecks.push({
  name: 'A29.1: rrweb/session.json contains > 0 events',
  expected: '>0',
  actual: events.length,
  passed: events.length > 0,
});
mergedChecks.push({
  name: `A29.2: rrweb emitted at least one Meta event (EventType.Meta=${EventType.Meta})`,
  expected: 'has Meta',
  actual: events.some((e) => e.type === EventType.Meta),
  passed: events.some((e) => e.type === EventType.Meta),
});
mergedChecks.push({
  name: `A29.3: rrweb emitted at least one FullSnapshot (EventType.FullSnapshot=${EventType.FullSnapshot})`,
  expected: 'has FullSnapshot',
  actual: events.some((e) => e.type === EventType.FullSnapshot),
  passed: events.some((e) => e.type === EventType.FullSnapshot),
});
mergedChecks.push({
  name: `A29.4: rrweb emitted at least one IncrementalSnapshot (EventType.IncrementalSnapshot=${EventType.IncrementalSnapshot})`,
  expected: 'has IncrementalSnapshot',
  actual: events.some((e) => e.type === EventType.IncrementalSnapshot),
  passed: events.some((e) => e.type === EventType.IncrementalSnapshot),
});

Use for: Plan 03-01 only. RESEARCH §"Pitfall 1" warns: synthetic probe HTML needs at least one DOM mutation between load and SAVE for IncrementalSnapshot to fire — Plan 03-01 driver MUST trigger one via page.evaluate(() => document.body.appendChild(...)) or page.click('#probe-modal-trigger').

UserEvent type-grep pattern (Plan 03-02) — derived from RESEARCH §"Code Examples" pattern 2 + src/shared/types.ts:124-131:

// Plan 03-02 driveA30 — events.json grep for 5 UserEvent.type values
import type { UserEvent } from '../../../src/shared/types';

const eventsRaw = await zip.file('logs/events.json')!.async('text');
const userEvents: UserEvent[] = JSON.parse(eventsRaw);

const EXPECTED_TYPES = ['click', 'input', 'navigation', 'js_error', 'network_error'] as const;
for (const expectedType of EXPECTED_TYPES) {
  mergedChecks.push({
    name: `A30.<i>: logs/events.json contains at least one '${expectedType}' event`,
    expected: `>=1 ${expectedType}`,
    actual: userEvents.filter((e) => e.type === expectedType).length,
    passed: userEvents.some((e) => e.type === expectedType),
  });
}

Negative-assertion pattern (Plan 03-03 password sentinel grep) — derived from RESEARCH §"Pattern 3" (no codebase analog; closest analog is driveA27's A27.7/A27.8 absence assertions at harness-page-driver.ts:1724-1737):

// Plan 03-03 driveA31 — sentinel absence from logs/events.json
const SENTINEL = 'secret-do-not-log-123';   // fixed; not a real secret
const eventsRaw = await zip.file('logs/events.json')!.async('text');
const userEvents: UserEvent[] = JSON.parse(eventsRaw);

const eventsContainingSentinel = userEvents.filter(
  (e) => e.value !== undefined && e.value.includes(SENTINEL),
);
mergedChecks.push({
  name: 'A31.1: password sentinel ABSENT from logs/events.json (existing src/content/index.ts:82 filter VERIFIED)',
  expected: 'absent (0 events containing sentinel)',
  actual: `${eventsContainingSentinel.length} events containing sentinel`,
  passed: eventsContainingSentinel.length === 0,
});

Filter-pipeline form (no continue) — driveA28 lines 1808-1810 (project style; CLAUDE.md "Control Flow" §):

const actualPaths: string[] = Object.keys(zip.files)
  .filter((path) => !zip.files[path].dir)
  .sort();

Use for: Any enumeration loop in Plans 03-01..04. No for...of + if (...) continue; shape.

Tab-cleanup safety pattern (T-02-04-04) — assertA27 lines 3271-3289 (try/catch silent-ignore on already-closed):

} finally {
  // T-02-04-04 mitigation: cleanup tabs with silent-ignore on already-closed.
  if (tabAId !== undefined) {
    try { await chrome.tabs.remove(tabAId); }
    catch (rmErr) { diag(result, `(tabA cleanup ignored: ${String(rmErr)})`); }
  }
  if (tabBId !== undefined) {
    try { await chrome.tabs.remove(tabBId); }
    catch (rmErr) { diag(result, `(tabB cleanup ignored: ${String(rmErr)})`); }
  }
}

Use for: Plan 03-02 if event-log trigger strategy involves opening a probe tab (e.g., https://example.com for js_error/network_error triggers).


tests/uat/harness.test.ts (orchestrator wiring)

Analog: tests/uat/harness.test.ts:65-130 + 316-335 + 406-430 (Plan 02-04)

Import section pattern — lines 92-100 (add A24-A28 imports per plan):

  // Plan 02-04 Task 1 — D-P2-01 empirical Blob URL verification
  driveA24,
  // Plan 02-04 Task 2 — REQ-archive-export-latency (5s ceiling)
  driveA25,
  // Plan 02-04 Task 3 — meta.json 8-field + multi-tab strict + REQ-archive-layout
  driveA26,
  driveA27,
  driveA28,
  getManifestVersion,
} from './lib/harness-page-driver';

Use for: Each Plan 03-01..04 appends ONE driveA<NN> import + a section banner comment naming the plan and assertion purpose.

Wrapped-driver pattern (when driver needs downloadsDir beyond the standard Page arg) — lines 322-335 (Plan 02-04 wrappers):

  // Plan 02-04 Task 2 — driveA25 needs downloadsDir for the host-side
  // dispatch→file-on-disk latency check (mirrors A5/A12/A13 wrapping).
  const driveA25Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
    (page) => driveA25(page, handles.downloadsDir);
  // Plan 02-04 Task 3 — driveA26/A27/A28 need downloadsDir for host-side
  // zip inspection (JSZip-parse meta.json + zip-layout enumeration). A26
  // chains off A25's zip (no new SAVE); A27 owns its SAVE (multi-tab);
  // A28 chains off A27's zip (no new SAVE).
  const driveA26Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
    (page) => driveA26(page, handles.downloadsDir);
  const driveA27Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
    (page) => driveA27(page, handles.downloadsDir);
  const driveA28Wrapped: (page: import('puppeteer').Page) => Promise<AssertionRecord> =
    (page) => driveA28(page, handles.downloadsDir);

Use for: Plans 03-01..03 (all need downloadsDir to find the latest zip). Same wrapping pattern; one wrapper per driver.

Drivers-array push pattern with explanatory banner comment — lines 397-430 (Plan 02-04):

    // Plan 02-04 Task 1 A24: D-P2-01 empirical Blob URL verification.
    // Installs chrome.downloads.onCreated listener cross-realm, dispatches
    // SAVE_ARCHIVE, captures the download URL, asserts the `blob:` prefix
    // (closes audit P0-6 end-to-end through a real Chrome instance +
    // the offscreen mint round-trip + chrome.downloads platform call).
    { name: 'A24', drive: driveA24 },
    // Plan 02-04 Task 2 A25: REQ-archive-export-latency / SPEC §10 #6.
    { name: 'A25', drive: driveA25Wrapped },
    // Plan 02-04 Task 3 A26: D-P2-02 + D-P2-03 meta.json 8-field shape.
    { name: 'A26', drive: driveA26Wrapped },
    // Plan 02-04 Task 3 A27: STRICT multi-tab urls[] post DEC-011 Amendment 1.
    { name: 'A27', drive: driveA27Wrapped },
    // Plan 02-04 Task 3 A28: REQ-archive-layout strict 5-entry zip-layout.
    { name: 'A28', drive: driveA28Wrapped },
  ];

Use for: Plans 03-01..04 each append ONE { name: 'A<NN>', drive: driveA<NN>Wrapped } entry with a banner comment citing the plan, the SPEC §10 # being verified, and the chaining strategy (own SAVE vs chain off prior).

FORBIDDEN_HOOK_STRINGS UAT A0 mirror — lines 115-130 (12 entries):

const FORBIDDEN_HOOK_STRINGS: ReadonlyArray<string> = [
  '__mokoshTest',
  'setCurrentStream',
  'setSegmentCountGetter',
  'installFakeDisplayMedia',
  'uninstallFakeDisplayMedia',
  'dispatchEndedOnTrack',
  'getSegmentCount',
  '__mokoshOffscreenQuery',
  'get-display-surface',
  'get-segment-count',
  // Plan 01-14 A23 surface — lockstep with unit-gate inventory at
  // tests/background/no-test-hooks-in-prod-bundle.test.ts:105.
  'lastGetDisplayMediaConstraints',
  'get-last-getDisplayMedia-constraints',
];

Lockstep requirement: ANY new test-only __MOKOSH_UAT__-gated symbol introduced by Plan 03-* MUST be added to BOTH this UAT A0 mirror AND the unit-gate at tests/background/no-test-hooks-in-prod-bundle.test.ts:108-126. RESEARCH §"Anti-Patterns" + A6 in Assumptions Log: Phase 3 plans are EXPECTED to stay at 12 entries (rrweb + chrome.downloads + chrome.tabs are production surfaces). The MEDIUM-risk exception (A6) is Plan 03-04 Page.metrics scaffolding — Page.metrics is host-side (not bundled into page-realm); no new symbol expected.


tests/uat/extension-page-harness.html (probe HTML append for Plan 03-01)

Analog: Current state of the file (lines 17-22) — append below, do NOT touch the head.

Existing scaffold (full file, 24 lines):

<!doctype html>
<html lang="en">
  <head>
    <meta charset="UTF-8">
    <title>Mokosh UAT Harness (extension-internal page)</title>
    <!--
      Plan 01-12 Wave 6: load the canonical token system on the harness
      page so A18 (Lora WOFF2 reachable via @font-face) and A21
      (--mks-font-display resolves to Lora) both have the @font-face
      rules + CSS custom properties + .mks-display-1 class visible via
      document.styleSheets + getComputedStyle. Vite's crxjs plugin
      handles the asset path rebasing at build:test time so the link
      resolves under dist-test/ even with content-hashed asset filenames.
    -->
    <link rel="stylesheet" href="../../src/shared/tokens.css">
  </head>
  <body>
    <h1>Mokosh UAT — extension-internal page harness</h1>
    <p>This page lives at <code>chrome-extension://&lt;id&gt;/tests/uat/extension-page-harness.html</code>.</p>
    <p>Puppeteer navigates a tab here and drives assertions via <code>window.__mokoshHarness.*</code>.</p>
    <pre id="status">Ready.</pre>
    <script type="module" src="./extension-page-harness.ts"></script>
  </body>
</html>

Probe-HTML append target: between <pre id="status"> (line 21) and <script type="module"> (line 22). Form + table + modal-trigger HTML per RESEARCH §"Pitfall 4" (NO <textarea>; rrweb alpha.4 bug #1596).

Critical constraints (per RESEARCH Assumption A2 + A5):

  • DO NOT modify the <head> block — <link rel="stylesheet" href="../../src/shared/tokens.css"> is load-bearing for A18 + A21.
  • DO NOT add stylesheet imports for the probe HTML — inline styles via style="..." or via the existing tokens.css --mks-* vars (defensive).
  • Probe HTML stays minimal: 1 form (<input type="text"> + <input type="email"> + <input type="password"> + <button>), 1 small table (2-3 rows), 1 hidden modal <div> toggled by a trigger button. Per RESEARCH §"Pitfall 1": at least one DOM mutation is required pre-SAVE for IncrementalSnapshot to fire.

.planning/phases/03-.../03-VERIFICATION.md (NEW; aggregator)

Analogs:

  • .planning/phases/02-stabilize-export-pipeline/02-VERIFICATION.md (T5 override pattern + overrides_applied field; closest match for Plan 03-05 charter)
  • .planning/phases/01-stabilize-video-pipeline/01-VERIFICATION.md (per-requirement scorecard + cross-cutting gates + operator empirical acks + deferred items)

Frontmatter shape (from 02-VERIFICATION.md:1-32):

---
phase: 02-stabilize-export-pipeline
verified: 2026-05-20T18:00:00Z
status: passed
score: 5/5 must-haves verified
overrides_applied: 1
override_notes:
  - dimension: "T5 operator empirical UAT (Plan 02-04 Task 4 Step 2)"
    initial_status: "UNCERTAIN (human_needed)"
    override_to: "VERIFIED"
    rationale: |
      User explicit delegation 2026-05-20: "why do i need to do all of this? It's on you to test..."
      established the operating principle that automation covers what automation can cover.
      Per saved memory feedback-trust-harness-over-manual-uat.md (created same session 2026-05-20),
      operator empirical UAT is reserved for surfaces automation genuinely cannot verify (brand
      judgment, ergonomics). For Phase 2 deep-pipeline work (Blob URL, meta schema, archive layout),
      every operator-checklist surface IS covered by harness assertions:
        - (a) <5s latency → A25 empirical via Puppeteer (real Chrome)
        - ... (5 sub-bullets covering each empirical channel)
      ... (citation block ending with "The harness IS the canonical Phase 2 verification per the current memory.")
human_verification: []
---

Phase 3 adaptation (per CONTEXT D-P3-02 + D-P3-04):

---
phase: 03-spec-10-smoke-verification-dom-event-log-verification
verified: 2026-05-20T<HH-MM-SS>Z
status: passed
score: 9/9 SPEC §10 criteria
overrides_applied: 1
override_notes:
  - dimension: "SPEC §10 #8 — Password masking (PARTIAL per D-P3-02 charter)"
    initial_status: "PARTIAL"
    rationale: |
      Full rrweb v2 maskInputFn + data-sensitive guards DEFERRED per 2026-05-20 charter
      'we don't care about privacy hardening. At least here.' REQ-password-confidentiality
      moved to Out of Scope v1. Existing src/content/index.ts:82
      `if (target.type === 'password') return;` filter VERIFIED by Plan 03-03 A31
      (sentinel string typed; greps events.json for absence).
human_verification:
  - dimension: "SPEC §10 #9 — Extension RAM ≤ 50 MB"
    rationale: |
      Per D-P3-04: operator instructions in VERIFICATION below — load extension, idle 5 min,
      open chrome://memory-internals OR chrome://extensions service-worker memory display,
      verify extension background RAM < 50 MB. Page.metrics scaffolding (if shipped by
      Plan 03-04) measures page-realm only and does not cover the SW target.
---

Per-Requirement Scorecard table (from 01-VERIFICATION.md:32-46):

## Per-Requirement Scorecard

| # | Requirement / Decision | Evidence | Status |
|---|---|---|---|
| 1 | REQ-video-ring-buffer | `src/offscreen/recorder.ts:52-58` (`MAX_SEGMENTS=3 × 10s = 30s`); ... | PASS |
| 2 | REQ-install-clean | `npm run build` exit 0, 9 chunks produced ... | PASS |
| ... |

Phase 3 adaptation: §10 #1-#9 row-by-row evidence table with Phase-citation (Phase 1 / Phase 2 / Phase 3) + plan-citation (e.g., "Plan 03-01 A29 GREEN") + commit-citation per criterion.

Cross-Cutting Gates table (from 01-VERIFICATION.md:48-56):

## Cross-Cutting Gates

| Gate | Evidence | Status |
|---|---|---|
| vitest | 28 files / **153 tests / 153 GREEN** ... | PASS |
| UAT harness | 24 drivers (A0 grep gate + A1..A14 + A15..A17 + A18..A22 + A23) ... | PASS |
| Tier-1 grep gate | 12 FORBIDDEN_HOOK_STRINGS ... | PASS |
| Pre-checkpoint bundle gates | 0 `googleapis`/`https://fonts` in dist; 0 test-hook leaks ... | PASS |
| tsc | `npx tsc --noEmit` exit 0 | PASS |

Phase 3 adaptation: Update UAT harness row (3X drivers post Plan 03-01..04); preserve baseline vitest + Tier-1 + bundle gates; cite Plan 03-05's own re-verification of all 6 pre-checkpoint bundle gates per saved memory feedback-pre-checkpoint-bundle-gates.md.

Operator-Empirical Acks table (from 01-VERIFICATION.md:70-78):

## Operator-Empirical Acks (verbatim + commit refs)

| Date | Plan | Operator response | Commit |
|---|---|---|---|
| 2026-05-15 | 01-07 (D-12 + A3 closure) | Chrome playback clean, ffmpeg dry-run exit 0 | cd61cbc |
| ... |

Phase 3 adaptation: Single row for the §10 #9 RAM operator ack (when received). Empty until ack lands; then verbatim quote + commit.

Deferred Items section (from 01-VERIFICATION.md:92-104 + 02-VERIFICATION.md:55-58):

## Forward-Looking Deferred Items (NOT gaps)

| Item | Owner | Source |
|---|---|---|
| rrweb v2 stable upgrade | Phase 4 | D-P3-03 defer rationale |
| Programmatic RAM measurement (puppeteer.Page.metrics enhancement OR chrome.devtools Memory API) | Phase 4 | D-P3-04 defer rationale |
| REQ-password-confidentiality v2 candidate (rrweb v2 maskInputFn + data-sensitive guards) | Phase 4 | D-P3-02 defer rationale |
| ... |

Phase 3 adaptation: Pull from CONTEXT §"Deferred" verbatim (10 items including the audit P1 polish + setimmediate polyfill + ROADMAP backfill).


tests/uat/lib/ramMetrics.ts (OPTIONAL — Plan 03-04 Page.metrics scaffolding)

Analog: NONE in codebase. Closest scaffolding shape: driveA25 host-side latency measurement pattern (tests/uat/lib/harness-page-driver.ts:1255-1334). New pattern; cite RESEARCH §"Code Example A3X" + §"Pitfall 2".

Decision per D-P3-04 + RESEARCH Open Question 3 (recommended: SHIP): Plan 03-04 ships ~30 lines of Page.metrics scaffolding either inline in harness-page-driver.ts (preferred per scope-minimization) OR as a new tests/uat/lib/ramMetrics.ts helper if planner prefers separation. RESEARCH §"Open Question 3" recommends shipping with explicit diagnostic copy.

Concrete pattern (per RESEARCH §"Code Example A3X"):

// Plan 03-04 optional driveA3X — Page.metrics() scaffolding (D-P3-04 best-effort)
async function driveA3X_OptionalRamScaffold(page: Page): Promise<AssertionRecord> {
  // Page.metrics is page-realm only — JSHeapUsedSize covers V8 isolate of THIS Page,
  // NOT the MV3 service worker (separate target).
  const metrics = await page.metrics();
  const jsHeapMB = metrics.JSHeapUsedSize !== undefined
    ? metrics.JSHeapUsedSize / (1024 * 1024)
    : -1;

  const checks: CheckRecord[] = [
    {
      name: 'A3X.1: Page.metrics returned JSHeapUsedSize',
      expected: '>= 0',
      actual: String(jsHeapMB),
      passed: jsHeapMB >= 0,
    },
    {
      name: 'A3X.2: Page-realm JS heap < 50 MB (NOTE: scaffolding only; SW context excluded)',
      expected: '< 50 MB',
      actual: `${jsHeapMB.toFixed(2)} MB`,
      passed: jsHeapMB < 50,
    },
  ];

  return {
    passed: checks.every((c) => c.passed),
    name: 'A3X — RAM scaffolding (best-effort; page-realm only per D-P3-04)',
    checks,
    diagnostics: [
      'NOTE: page-realm only; SW context measurement requires chrome://memory-internals operator verification per D-P3-04.',
    ],
  };
}

Critical diagnostic copy gate (per RESEARCH §"Pitfall 2"): The scaffolding MUST emit 'NOTE: page-realm only; SW context excluded' in diagnostics so an operator does not interpret a green automation number as full §10 #9 closure. The binding §10 #9 gate is the operator chrome://memory-internals check; scaffolding is informational only.


Shared Patterns

Approach B Harness Extension Lockstep

Source: Plan 02-04 + Plan 01-13 (Approach B foundation) Apply to: Plans 03-01, 03-02, 03-03, 03-04 (every Phase 3 plan that ships a new assertion)

The 3-file lockstep across each new assertion:

  1. tests/uat/extension-page-harness.ts — append async function assertA<NN>(): Promise<AssertionResult> (page-side stub OR orchestrator) + extend declare global interface + add entry to window.__mokoshHarness = {...} object literal.
  2. tests/uat/lib/harness-page-driver.ts — append export async function driveA<NN>(page: Page, downloadsDir: string): Promise<AssertionRecord> (3-phase pattern: page.evaluate stub → findLatestZip → JSZip read).
  3. tests/uat/harness.test.ts — append driveA<NN> import + wrapped-driver const (if downloadsDir is needed) + drivers-array push with banner comment.

Missing ANY one of the three breaks the orchestrator (either at the harness.assertA<NN>() page.evaluate call OR at the for (const drive of drivers) loop). Plan-checker MUST validate all three are touched per new A-number.

Pre-Checkpoint Bundle Gates (6/6 Standard Inventory)

Source: Saved memory feedback-pre-checkpoint-bundle-gates.md + 02-04-SUMMARY.md:158-168 (canonical 7-gate table; Phase 2 closure precedent) Apply to: Plan 03-05 (before VERIFICATION.md is written) — ALL plans before any operator step

Standard 6/6 inventory (Plan 02-04 ran 7; Plan 03-05 should run the same):

Gate Concrete Check
Gate 1 npm run build exit 0
Gate 2 SW CSP-safety: grep -rn "new Function\|eval(" dist/assets/ — 1 documented exception (setimmediate polyfill)
Gate 3 SW Node-globals: grep -rn "Buffer.from\|Buffer.alloc\|require(" dist/assets/index.ts-*.js — 0 hits
Gate 4 DOM-globals: grep -rn "window.\|document." dist/assets/index.ts-*.js — bundled-lib idiom (typeof-guarded)
Gate 5 Tier-1 SW-bundle-import gate (tests/background/sw-bundle-import.test.ts) GREEN
Gate 6 FORBIDDEN_HOOK_STRINGS unit gate (tests/background/no-test-hooks-in-prod-bundle.test.ts) — 12 strings, 0 hits each
Gate 7 Manifest validation gates: i18n + locale-parity + build — all GREEN

T5 Override Pattern (Harness-Coverage-as-Verification)

Source: 02-VERIFICATION.md:1-31 + saved memory feedback-trust-harness-over-manual-uat.md Apply to: Plan 03-05 VERIFICATION.md — for §10 criteria where harness coverage exists (§10 #4 / #5 / #8) but NOT for §10 #9 RAM (genuine non-automatable)

When verifier returns human_needed for a criterion AND a harness assertion empirically covers the same surface:

  • Move human_verification entry to overrides_applied block in frontmatter
  • Include explicit user-delegation citation (date + verbatim quote, if recorded) + saved-memory reference
  • Sub-bullet each harness assertion that covers the surface
  • Mark status: passed rather than status: human_needed

Counter-example (do NOT override): §10 #9 RAM measurement — Page.metrics scaffolding is page-realm only (RESEARCH §"Pitfall 2"); the SW context can ONLY be measured via operator-driven chrome://memory-internals. This is the canonical human_verification entry per D-P3-04.

Filter-Pipeline Form (No continue)

Source: ~/.claude/CLAUDE.md "Control Flow" § + Plan 02-04 driveA28 (lines 1808-1810) Apply to: Any enumeration loop in Plans 03-01..04

// PREFERRED — filter pipeline
const matchingEvents = userEvents.filter((e) => e.type === 'click');
const realHttpUrls = urls.filter((u) => /^https?:\/\//.test(u));

// AVOID — for...of + continue
// for (const e of userEvents) { if (e.type !== 'click') continue; ... }

Production-Surface-Only New Assertions (FORBIDDEN_HOOK_STRINGS stays at 12)

Source: 02-04-SUMMARY.md:99 + RESEARCH Assumption A6 Apply to: Plans 03-01..04 — riding production surfaces (rrweb.record + content-script GET_RRWEB_EVENTS + chrome.tabs.create + chrome.downloads.onCreated)

Phase 3 assertions ride production surfaces verified by Plans 01-13 + 02-04. NO new __MOKOSH_UAT__-gated test-only symbols expected. If Plan 03-04 Page.metrics scaffolding requires a new bridge op (e.g., get-page-metrics from offscreen → harness), that scenario adds 1-2 entries to BOTH FORBIDDEN_HOOK_STRINGS inventories AND requires Vite define-token tree-shake gating. Plan-checker MUST validate inventory count.

Anti-Pattern: Re-reading the Same Range

Source: CLAUDE.md operator guidance Apply to: All Phase 3 implementation work — read each source file ONCE, extract patterns, then implement


No Analog Found

File Role Data Flow Reason / Mitigation
tests/uat/lib/ramMetrics.ts (if Plan 03-04 chooses separate-file scaffolding) test (host-side utility) request-response (CDP Performance.getMetrics) Page.metrics() has no codebase precedent. RESEARCH §"Code Example A3X" provides the canonical implementation; planner uses verbatim. Plan 03-04 may inline this in harness-page-driver.ts instead of new file (scope-minimization preferred).

All other anticipated files have direct analogs from Plan 02-04 (most recent Approach B precedent).

Per-Plan Anticipated File Touch Inventory

Plan Files Modified Files Created Pattern Source
03-01 (rrweb DOM §10 #4) tests/uat/extension-page-harness.ts + tests/uat/extension-page-harness.html (probe HTML) + tests/uat/lib/harness-page-driver.ts + tests/uat/harness.test.ts Plan 02-04 (assertA26/driveA26 + JSZip + EventType import)
03-02 (event-log §10 #5) tests/uat/extension-page-harness.ts + tests/uat/lib/harness-page-driver.ts + tests/uat/harness.test.ts Plan 02-04 (driveA26 chain pattern + UserEvent grep from src/shared/types.ts:124-131)
03-03 (password-filter §10 #8 PARTIAL) tests/uat/extension-page-harness.ts + tests/uat/lib/harness-page-driver.ts + tests/uat/harness.test.ts (possibly probe HTML if Plan 03-01 didn't add <input type="password">) RESEARCH §"Pattern 3" + driveA27 absence-assertion shape
03-04 (RAM §10 #9 best-effort) OPTIONAL: tests/uat/lib/harness-page-driver.ts (Page.metrics inline) + tests/uat/harness.test.ts (wrapped driver) OPTIONAL: tests/uat/lib/ramMetrics.ts RESEARCH §"Code Example A3X" (no codebase analog)
03-05 (§10 sweep aggregator) .planning/phases/03-.../03-VERIFICATION.md 01-VERIFICATION.md (scorecard structure) + 02-VERIFICATION.md (T5 override frontmatter)

Wave Sequencing Note

Per CONTEXT §"Claude's Discretion" + RESEARCH Open Question 1 + Pitfall 6: Plans 03-01..04 modify the SAME three harness files (extension-page-harness.ts, harness-page-driver.ts, harness.test.ts). RESEARCH recommends SEQUENTIAL execution within Wave 2: 03-01 → 03-02 → 03-03 → 03-04. Plan-checker validates files_modified overlap audit. Plan 03-05 (VERIFICATION.md aggregator) runs in Wave 3 after 03-01..04 land + UAT harness reaches 33/33 (or 32/32 if Plan 03-04 skips scaffolding).

Metadata

Analog search scope: tests/uat/, .planning/phases/01-stabilize-video-pipeline/, .planning/phases/02-stabilize-export-pipeline/, src/content/index.ts, src/shared/types.ts, node_modules/@rrweb/types/dist/index.d.ts

Files read (each ONCE, non-overlapping ranges):

  • tests/uat/extension-page-harness.ts:2840-3409 (assertA24-A28 + __mokoshHarness wire)
  • tests/uat/lib/harness-page-driver.ts:1-130, 1180-1349, 1360-1599, 1700-1849 (4 non-overlapping ranges; covers driveA24/A25/A26/A27/A28 + findLatestZip + A28_EXPECTED_PATHS + imports)
  • tests/uat/harness.test.ts:1-100, 100-260, 320-440 (3 non-overlapping ranges; covers orchestrator imports + FORBIDDEN_HOOK_STRINGS + A0 gate + drivers wrap + push)
  • tests/uat/extension-page-harness.html (full; 24 lines)
  • tests/uat/lib/zip.ts (full; 140 lines)
  • tests/background/no-test-hooks-in-prod-bundle.test.ts:95-175 (FORBIDDEN_HOOK_STRINGS unit-gate)
  • .planning/phases/01-stabilize-video-pipeline/01-VERIFICATION.md (full; 123 lines)
  • .planning/phases/02-stabilize-export-pipeline/02-VERIFICATION.md (full; 166 lines)
  • .planning/phases/02-stabilize-export-pipeline/02-04-SUMMARY.md (full; 259 lines)
  • src/content/index.ts:75-89 (password filter at line 82) + grep for rrweb wiring anchors
  • src/shared/types.ts:120-155 (UserEvent + SessionMetadata)
  • node_modules/@rrweb/types/dist/index.d.ts:186-194 (EventType enum)

Pattern extraction date: 2026-05-20