visual-qa-dogfood

Verifies UI changes with screenshot evidence and an absurdity ledger of visual defects.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/Lawrence908/chiron --skill visual-qa-dogfood-lawrence908
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: visual-qa-dogfood
Source: https://github.com/Lawrence908/chiron/tree/main/plugins/dev/skills/visual-qa-dogfood
Command: npx skills add https://github.com/Lawrence908/chiron --skill visual-qa-dogfood-lawrence908

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Automated tests and DOM metrics can pass while a rendered page still looks broken to a human. This Skill closes that gap by forcing screenshot-backed inspection of every user-visible surface a change touches before the work is called done. ## Core Features & Use Cases - Screenshot-backed verification: Captures every touched surface across desktop and narrow viewports, themes, and empty/loading/error states using tools like the Playwright MCP. - Absurdity ledger: Records each visual defect with screenshot path, viewport, user impact, and a disposition (ready, fixed, intentionally accepted, out of scope, needs review). - Fix-and-recapture loop: Patches the highest-signal defect, then re-captures evidence from the fixed build, with an optional subagent review gate for non-trivial changes. - Use Case: After refactoring a Next.js dashboard layout, render the affected routes at 1920x1080 and mobile widths, log a clipped sidebar and an unreadable contrast issue in the ledger, fix both, and attach fresh screenshots as proof. ## Quick Start Use the visual-qa-dogfood skill to screenshot every surface touched by this diff and report any visual defects before we call it done.

Frequently Asked Questions about visual-qa-dogfood

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I visually verify a frontend change beyond passing tests?▼

Render every surface the diff touches and capture screenshots at realistic viewports, such as 1920x1080 desktop plus a narrow width. Inspect each capture like a user would, log defects in an absurdity ledger, and re-capture evidence after fixes.

What tool captures screenshots for visual QA of web apps?▼

The Playwright MCP is the usual capture tool, using browser_navigate followed by browser_take_screenshot against a dev server or a container reachable through your reverse proxy. Any rendering path that produces real screenshots of the affected surfaces works.

When should I use visual QA instead of functional bug hunting?▼

Use visual QA when a diff touches UI, CSS, layout, typography, or rendered output and you need to confirm it looks right to a human. Functional bugs belong to a bug-hunt skill and code quality to code review; this only asks whether it looks correct.

Can I test layouts with small placeholder content?▼

No. A fixture with three words does not exercise the layout. Use the user's own screenshot or source fixture plus at least one dense, real-world example, and never edit live user data to create a test case.

Why is a passing probe not enough to declare a UI task done?▼

Probes report dimensions, clipping, and contrast metrics but cannot judge whether a page looks broken to a human. Done requires visual evidence of touched surfaces, no unresolved ledger items, passing automated checks, and verification of the deployed surface.