Borrowing it
Nothing to install: this file belongs to HomenShum/NodeBenchAI. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/HomenShum/NodeBenchAI/main/.claude/skills/live-qa-dogfood-loop/SKILL.mdgit clone --depth 1 https://github.com/HomenShum/NodeBenchAIWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/homenshum/nodebenchai/live-qa-dogfood-loop)<a href="https://agentmods.dev/skills/homenshum/nodebenchai/live-qa-dogfood-loop"><img src="https://agentmods.dev/badge/skills/homenshum/nodebenchai/live-qa-dogfood-loop.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00017 | $0.01452 |
| Opus 5 | $0.00009 | $0.00726 |
| Sonnet 5 | $0.00003 | $0.00290 |
| Haiku 4.5 | $0.00002 | $0.00145 |
Grade A, and why
live-qa-dogfood-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 149 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Live QA Dogfood Loop
Self-conducting improvement loop that uses Gemini 3 Pro vision to screenshot, judge, fix, and re-judge all 5 app surfaces until top-market quality.
Quick start
# Ensure dev server is running on port 5191
npm run dev
# Run the QA judge (single pass)
GEMINI_API_KEY=$(grep GEMINI_API_KEY .env.local | cut -d= -f2) \
node scripts/ui/liveQaJudge.mjs --baseURL=http://127.0.0.1:5191
# Results land in .tmp/live-qa-judge/<timestamp>/
# History tracked in .tmp/live-qa-history.json
What the judge evaluates
8 screenshot captures per run:
- home —
/?surface=home(desktop 1440x900) - home-interaction — after clicking first card on home
- chat —
/?surface=chatwith a pre-loaded query + lens - reports —
/?surface=reports - reports-interaction — after clicking first report card
- nudges —
/?surface=nudges - me —
/?surface=me - mobile-home —
/?surface=homeat 375x812
Each screenshot is sent to Gemini 3 Pro (gemini-3-pro-preview) with a structured prompt requesting:
- Per-surface score (1-10)
- Strengths (what works)
- Issues with severity (P1/P2/P3), category, description, fix, and element
Scoring
Grade S: 95+ (top market quality)
Grade A: 90-94
Grade B: 80-89
Grade C: 70-79
Grade D: 60-69
Self-improvement loop protocol
Phase 1: Run judge, capture baseline
node scripts/ui/liveQaJudge.mjs --baseURL=http://127.0.0.1:5191
Phase 2: Fix P1s first (each P1 = ~6 points)
- Read
.tmp/live-qa-judge/<latest>/gemini-judge-response.json - For each P1: trace root cause (not symptom), fix the cause
- Common P1 patterns:
- Toast overlap — adjust Sonner position or z-index in
src/main.tsx - Unreadable text — increase contrast:
--rd-ink-faint->--rd-ink-soft - Broken layout — check CSS grid/flex alignment
- Toast overlap — adjust Sonner position or z-index in
Phase 3: Fix P2s (each P2 = ~2 points)
- Low contrast — darken
--text-muted/ brighten--rd-ink-faintin dark mode - Badge urgency — increase border opacity + add subtle glow for act-now badges
- Mobile viewport — compact or make dismissible promo cards
- Button states — distinct disabled (opacity 0.35 + grayscale) vs active states
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 149 lines · 17 tokens per session scan A c6b36137c07b
live-qa-dogfood-loop is a skill published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed yesterday), licensed MIT. It adds 17 tokens to every session and 1,452 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
screen-reader-testing
Test web applications with screen readers including VoiceOver, NVDA, and JAWS. Use when validating screen reader compatibility, debugging accessibility issues, or ensuring assistive technology support.
e2e-testing-patterns
Master end-to-end testing with Playwright and Cypress to build reliable test suites that catch bugs, improve confidence, and enable fast deployment. Use when implementing E2E tests, debugging flaky tests, or establishing testing standards.
playwright-cli
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
dogfood
Systematically explore and test a mobile app on iOS/Android with agent-device to find bugs, UX issues, and other problems. Use when asked to dogfood, QA, exploratory test, find issues, bug hunt, or test this app on mobile.
relay-80-100-workflow
Use when writing agent-relay workflows that must fully validate features end-to-end before merging. Covers the 80-to-100 pattern - going beyond "code compiles" to "feature works, tested E2E locally." Includes repair-before-failure validation gates, mandatory sequential Claude-then-Codex fresh-eyes review/fix loops…
browser-testing-with-screenshots
Use when testing web applications with visual verification - automates Chrome browser interactions, element selection, and screenshot capture for confirming UI functionality.