Skip to content

Quality Assurance

MissionChief UK uses layered validation so structured data, generated exports, documentation and interactive tools are checked before and after deployment.

Browser coverage

The Playwright acceptance suite runs against:

Project Engine Operational target
chromium-desktop Chromium Desktop Chrome-class browsers
firefox-desktop Firefox Desktop Firefox
webkit-iphone WebKit iPhone-sized iOS Safari behaviour
webkit-ipad WebKit iPad-sized Safari behaviour

Playwright device profiles emulate viewport, touch, mobile mode and user-agent behaviour. They do not replace occasional checks on physical Apple hardware.

Acceptance coverage

The suite verifies:

  • Command Centre and every public intelligence-tool route;
  • generated FAQ, API guide and release notes;
  • absence of JavaScript exceptions and first-party HTTP failures;
  • responsive content without page-level horizontal overflow;
  • mission lookup loading and filtering;
  • resource and qualification comparison;
  • concurrent fleet multiplication;
  • deterministic query-catalogue results;
  • the global command palette opening through Ctrl+K or ⌘K;
  • command-palette search against the generated canonical search index;
  • collection filtering, keyboard closure and mission-result deep linking;
  • iPhone viewport operation of the command palette;
  • MkDocs Material instant-navigation reinitialisation;
  • all eight static API endpoints and their cross-file counts;
  • critical WCAG A/AA violations on the main interactive surfaces.

The deterministic documentation audit additionally checks:

  • relative Markdown links;
  • local HTML href and src references;
  • GitHub Pages URLs that point back into this project;
  • MkDocs page-route resolution for both page.md and page/index.md layouts;
  • local heading anchors in README, changelog and documentation pages.

External websites are not treated as deterministic build dependencies because availability, anti-bot controls and redirects can change independently of this repository. Official evidence URLs remain recorded in the production data and are reviewed through the evidence-maintenance process.

Validation layers

Canonical JSON records
        ↓
Schema and referential validation
        ↓
Generated exports and FAQ
        ↓
Repository/API readiness audit
        ↓
Documentation link and anchor audit
        ↓
Strict MkDocs build
        ↓
All JavaScript syntax validation
        ↓
Chromium acceptance against the built site
        ↓
Command-palette, deep-link and mobile acceptance
        ↓
GitHub Pages deployment
        ↓
HTTP/API smoke test
        ↓
Chromium, Firefox, iPhone WebKit and iPad WebKit acceptance

A failed post-deployment browser test marks the Pages workflow as failed and prevents automated release publication.

Command-search trust boundary

The global command palette is tested as a read-only interface over assets/data/v1/search-index.json.

It must:

  • load only the generated first-party search index;
  • expose no arbitrary HTML from data records;
  • escape record content before rendering;
  • avoid MissionChief authentication and account access;
  • preserve omitted values as unknown;
  • link mission records into Mission Lookup using an encoded query; and
  • link other collections into the Query Catalogue using the same encoded-query contract.

Failure diagnostics

Failed browser runs upload an Actions artifact containing:

playwright-report/
test-results/

Depending on the failure, this can include:

  • HTML test report;
  • Playwright trace;
  • screenshot;
  • retained video;
  • console and network failure details.

Artifacts are retained for 14 days.

Link-audit failures are printed directly in the Actions log using this format:

source-file.md:line: missing local target or anchor

Local execution

Run the deterministic repository checks:

python scripts/validate_data.py
python scripts/generate_exports.py
python scripts/generate_faq.py
python scripts/release_readiness.py
python scripts/audit_links.py
mkdocs build --strict --site-dir site
python scripts/release_readiness.py --site-dir site

Validate every browser-side script:

for file in docs/javascripts/*.js; do node --check "$file"; done

Install browser-test dependencies and browser binaries:

npm install
npx playwright install --with-deps

Run against the public site:

npm run test:e2e

Run one browser project:

npm run test:e2e -- --project=chromium-desktop

Run only the command-palette acceptance file:

npm run test:e2e -- tests/e2e/command-palette.spec.mjs

Run against another deployment or local server:

MCUK_BASE_URL=http://127.0.0.1:8000/ npm run test:e2e

On Windows PowerShell:

$env:MCUK_BASE_URL = "http://127.0.0.1:8000/"
npm run test:e2e

Scope boundary

Automated tests can detect repeatable browser, accessibility, network and data-contract defects. They cannot fully reproduce:

  • every physical iPhone or iPad model;
  • assistive-technology combinations;
  • unusual browser extensions or content blockers;
  • poor mobile networks;
  • subjective readability and visual-design preferences.