away-mission-qa

name	away-mission-qa
description	Structured browser QA with health scoring (0-100), fix cycles, and regression tests. Use when testing a deployed site, running QA, or user says 'away mission qa' or 'test the site'.

/away-mission-qa -- Structured Browser QA

Automated QA mission against a live or staging URL. Produces a health score (0-100), runs fix cycles for issues found, and tracks regression between runs.

Arguments

• /away-mission-qa <url> -- full QA sweep of target URL
• /away-mission-qa <url> --quick -- critical checks only (auth, 404s, console errors)
• /away-mission-qa <url> --regression -- compare against previous run
• /away-mission-qa <url> --fix -- auto-fix mode: attempt fixes and re-test (max 3 cycles)
• /away-mission-qa <url> --checklist <path> -- use custom checklist file

Prerequisites

• Playwright MCP available (project-scoped .mcp.json or global)
• Target URL must be reachable from the current environment
• For --fix mode: write access to the project source code

Health Score Formula

Score starts at 100, deductions per category:

Category	Weight	Checks	Deduction per fail
Critical Errors	30	Console errors, JS exceptions, 4xx/5xx responses	-10 each (cap -30)
Navigation	20	Broken links, redirect chains, 404 pages	-5 each (cap -20)
Visual	15	Layout shifts, overflow, missing images, contrast	-3 each (cap -15)
Forms	15	Validation, submission, CSRF, error states	-5 each (cap -15)
Performance	10	LCP > 2.5s, CLS > 0.1, TBT > 200ms	-3 each (cap -10)
Accessibility	10	Missing alt text, unlabeled inputs, color contrast	-2 each (cap -10)

Thresholds: 90-100 GREEN, 70-89 AMBER, 0-69 RED

Workflow

flowchart TD
    Start([away-mission-qa]) --> Parse["Parse URL + flags"]
    Parse --> Quick{--quick?}
    Quick -- Yes --> Critical["Run critical checks only\n(console, status codes, auth)"]
    Quick -- No --> Full["Full QA sweep\n(all 6 categories)"]
    Critical --> Score["Calculate health score"]
    Full --> Score
    Score --> Report["Generate QA report"]
    Report --> Regress{--regression?}
    Regress -- Yes --> Compare["Compare with previous run\nin .qa/reports/"]
    Regress -- No --> FixMode{--fix?}
    Compare --> FixMode
    FixMode -- Yes --> Cycle["Fix cycle 1/3:\n- identify fixable issues\n- apply fixes\n- re-run failed checks"]
    FixMode -- No --> Save["Save report to .qa/reports/"]
    Cycle --> Improved{Score improved?}
    Improved -- "Yes + cycles < 3" --> Cycle
    Improved -- "No or cycles = 3" --> Save
    Save --> Done([Done])

QA Check Details

Critical Errors (weight: 30)

Using Playwright MCP or Bash with curl:

1. Console errors: Navigate to each page, capture console.error events
2. HTTP status codes: Check all pages return 200 (or expected redirect)
3. JavaScript exceptions: Capture uncaught exceptions during page load
4. API failures: Monitor network requests for 4xx/5xx responses
5. Auth flow: If login page exists, verify login/logout cycle works

Navigation (weight: 20)

1. Internal links: Crawl all internal links from homepage (max 50 pages)
2. Broken links: Flag any link returning 404 or connection refused
3. Redirect chains: Flag chains > 2 hops
4. Anchor links: Verify #fragment targets exist on page

Visual (weight: 15)

Using Playwright screenshots + visual inspection:

1. Layout overflow: Check for horizontal scrollbar on standard viewports (375px, 768px, 1280px)
2. Missing images: Find <img> tags with broken src
3. CLS: Measure cumulative layout shift during page load
4. Text contrast: Check WCAG AA contrast ratios on visible text

Forms (weight: 15)

1. Empty submission: Submit forms without filling required fields, verify error messages
2. Validation messages: Check client-side validation fires correctly
3. Success state: Submit with valid data, verify success response
4. CSRF protection: Check for CSRF token presence in forms

Performance (weight: 10)

Using Lighthouse-style metrics (via Playwright performance API):

1. LCP (Largest Contentful Paint): threshold 2.5s
2. CLS (Cumulative Layout Shift): threshold 0.1
3. TBT (Total Blocking Time): threshold 200ms

Accessibility (weight: 10)

1. Image alt text: All <img> have alt attributes
2. Form labels: All inputs have associated <label> or aria-label
3. Heading hierarchy: No skipped heading levels (h1 -> h3)
4. Keyboard navigation: Tab order reaches all interactive elements

Report Format

## QA Report -- {url}

**Date**: {YYYY-MM-DD HH:mm}
**Health Score**: {score}/100 ({GREEN|AMBER|RED})
**Pages tested**: {count}
**Duration**: {seconds}s

### Score Breakdown

| Category | Score | Issues |
|----------|-------|--------|
| Critical Errors | {x}/30 | {count} |
| Navigation | {x}/20 | {count} |
| Visual | {x}/15 | {count} |
| Forms | {x}/15 | {count} |
| Performance | {x}/10 | {count} |
| Accessibility | {x}/10 | {count} |

### Issues Found

#### Critical
- [C-001] {description} -- {page URL}

#### Warnings
- [W-001] {description} -- {page URL}

#### Suggestions
- [S-001] {description} -- {page URL}

### Regression (if --regression)
| Metric | Previous | Current | Delta |
|--------|----------|---------|-------|
| Health Score | {prev} | {curr} | {+/-} |
| Critical Issues | {prev} | {curr} | {+/-} |

### Fix Cycles (if --fix)
| Cycle | Issues Fixed | Score Before | Score After |
|-------|-------------|-------------|-------------|
| 1 | {count} | {before} | {after} |

Persistence

• Reports saved to .qa/reports/{domain}-{YYYY-MM-DD-HHmm}.md
• Previous reports used for --regression comparison (most recent matching domain)
• Create .qa/ directory if it does not exist

Custom Checklist

When --checklist <path> is provided, load a markdown file with custom checks:

# Custom QA Checklist

- [ ] Homepage hero loads correctly
- [ ] Contact form submits without error
- [ ] Pricing page shows correct plans
- [ ] Blog posts render markdown properly

Each item is verified manually via Playwright navigation and added to the report.

Worked Example

User: /away-mission-qa https://staging.example.com --fix

1. Preflight: URL reachable (200), Playwright available
2. Full sweep: 12 pages crawled, 6 categories tested
3. Score: 73/100 (AMBER)
   • Critical: 30/30 (no errors)
   • Navigation: 15/20 (1 broken link)
   • Visual: 12/15 (1 missing image)
   • Forms: 10/15 (contact form missing validation)
   • Performance: 4/10 (LCP 3.2s, CLS 0.15)
   • Accessibility: 2/10 (4 images missing alt, 2 unlabeled inputs)
4. Fix cycle 1: Fixed broken link href, added missing alt texts, added form validation
5. Re-test: Score improved to 88/100 (AMBER)
6. Fix cycle 2: Optimized hero image (LCP), fixed CLS with explicit dimensions
7. Re-test: Score improved to 94/100 (GREEN)
8. Report saved: .qa/reports/staging.example.com-2026-03-23-1430.md

Output: "Away mission complete. Health score: 94/100 (GREEN). 2 fix cycles applied. Report saved."

Troubleshooting

Symptom	Cause	Fix
"Playwright MCP not available"	Not configured for this project	Add Playwright to .mcp.json or use --curl-only for HTTP-only checks
Score stuck at 0	Site not reachable or returning 5xx	Verify URL is accessible; check VPN/firewall
Fix cycle makes no progress	Issues require manual intervention	Review report, fix manually, re-run with --regression
Regression comparison fails	No previous report found	Run a baseline first without --regression
Performance metrics unavailable	Playwright performance API not supported	Deduction skipped; score calculated from remaining categories
"Max pages reached"	Site has > 50 internal links	Expected; crawl limited to 50 pages for efficiency