verify-ui-visually
Skill98 starsChanged 2 months agoMITView on GitHub
Verify UI changes by rendering and looking — screenshot, browser, or running app — never by reading the markup and imagining it. Use after any change to components, styles, CSS, layouts, templates, charts, emails, PDFs, or generated documents/images, and when a user says "it looks wrong", "check the page", or asks for a visual fix. Not for pure logic changes with no rendered output.
In this repository38 other files it gives its agents.
- ambiguity-commitSkill
- api-surface-careSkill
- calibrated-recommendationSkill
- codebase-orientationSkill
- concurrency-reasoningSkill
- config-and-flagsSkill
- context-checkpointSkill
- cross-platform-careSkill
- data-loss-guardSkill
- data-migration-safetySkill
- dependency-changesSkill
- docs-as-contractSkill
- edge-case-sweepSkill
- environment-firstSkill
- error-message-qualitySkill
- escalate-vs-decideSkill
- estimation-and-scopingSkill
- explain-at-the-right-levelSkill
- generalize-the-correctionSkill
- git-hygieneSkill
- incident-diagnosisSkill
- leave-no-messSkill
- llm-app-codeSkill
- numerical-careSkill
- performance-investigationSkill
- perf-sanitySkill
- question-vs-taskSkill
- resource-lifecycleSkill
- reviewing-a-changeSkill
- root-cause-debuggingSkill
- security-reflexesSkill
- stop-thrashingSkill
- subagent-fanoutSkill
- surgical-refactoringSkill
- task-decompositionSkill
- test-designSkill
- untrusted-content-guardSkill
- workmanshipSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Capture a baseline first: Before editing, screenshot the affected screen (or its unchanged sibling) at the viewport you will test.
- Very specific: line 11
Render it and look at it: Run the dev server and take a screenshot (browser tooling, mcp__chrome-devtools__take_screenshot, or the …
- Distinctive: line 15
Check both themes and at least two widths: when the project supports them: light/dark, mobile/desktop breakpoint.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
verify-ui-visually
Weak UI verification looks like: edit the markup, run the type-check, imagine the result, report "done". Strong verification puts pixels in front of your eyes and reports what they show. Code that compiles tells you nothing about what the user sees.
- Capture a baseline first. Before editing, screenshot the affected screen (or its unchanged sibling) at the viewport you will test. Without a before, you cannot tell a regression you introduced from a quirk that was already there.
- Render it and look at it. Run the dev server and take a screenshot (browser tooling,
mcp__chrome-devtools__take_screenshot, or the project's snapshot setup); open the generated PDF/image/email HTML in a real viewer. Reading the diff is not seeing the output. - Confirm you are looking at your change. Find a string or selector from your diff in the rendered DOM or page text before judging pixels. An unchanged screenshot means a stale build, cached page, or wrong route until proven otherwise — fix that before concluding the change is a no-op.
- Look at the actual pixels critically: overflow and truncation, misalignment, spacing collapse, z-index sandwiches, contrast, loading and empty states — the bugs no test asserts and every user notices instantly.
- Check the states, not just the happy render: empty (zero items), loaded (typical), overloaded (long strings, many items — "WWWWWWWW" and a 200-char title), error, and loading. Most layout breaks live in empty and overloaded.
- Check both themes and at least two widths when the project supports them: light/dark, mobile/desktop breakpoint. A change verified only in dark-mode desktop ships broken for half the users.
- Interact, don't just render: click the button you changed, submit the form, tab through for focus order. Console errors during the interaction count as failures even when the screen looks fine.
- When you can't render (no display, missing env), say so explicitly and downgrade the claim: "markup changed as requested; not visually verified — needs a look at /settings in the browser."
Report format: name the screenshot files, the viewport(s) and theme(s) checked, and describe in words what you saw ("badge now sits on the baseline; long titles truncate with an ellipsis at 320px"). "Screenshot taken, looks good" is not a verification — it is the claim you were supposed to check.
Rule: a UI change reported as done carries either a screenshot-verified description or a stated inability to verify. Never the silent middle.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

