agentleFS
jazzpujols34jazzpujols34/claude-skills/skills/verify-frontend-change/SKILL.md

verify-frontend-change

Skill0 starsChanged 5 months agoMITView on GitHub

Verify any UI change end-to-end before declaring it done. Trigger after any change to a component, page, style, layout, or interactive control — and whenever you're about to say a frontend task is "done", "fixed", "working", or "complete". Auto-activate when you edited frontend code and are tempted to report success from a clean edit/typecheck alone.

2 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Run these in order. Do not skip ahead.

    line 26
  2. Distinctive:

    Start the dev server and open the edited page — not the home page, the actual route you touched.

    line 30
  3. Distinctive:

    Check dark mode and mobile width if the change touches layout, color, or spacing.

    line 46
  4. Distinctive:

    If any step fails, fix the issue and rerun from step 1.

    line 64
  5. Distinctive:

    If any row is blank, the task is not done.

    line 80
  6. Distinctive:

    Screenshots are the deliverable, not decoration: When you report the change as verified, attach or reference the before/after so the claim …

    line 101

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Skill: Verify Frontend Change

Never report a UI change as complete based on a successful edit alone. A green typecheck proves the code compiles, not that the pixel moved, the button fires, or the console stayed clean. Verify it the way a human reviewer would — in a real browser.

Detection — When to Activate

Activate this skill when:

  • You edited a component, page, route, style, layout, or interactive control
  • You're about to say a frontend task is "done", "fixed", "working", or "shipped"
  • You added or changed a control (button, input, toggle, modal, dropdown, form)
  • A change is theme-, responsive-, or state-dependent (dark mode, mobile, loading/empty/error)

Do NOT activate for:

  • Pure backend / API / data-layer changes with no rendered surface
  • Docs, comments, config, or type-only changes that don't alter output
  • Copy-only edits already confirmed by reading the diff (still glance at the rendered page)

The Protocol

Run these in order. Do not skip ahead.

1. Open the change in a real browser

Start the dev server and open the edited page — not the home page, the actual route you touched.

npm run dev        # or the project's equivalent (Vite / Next / etc.)
# then open the specific path you changed, e.g. http://localhost:3000/settings

Confirm the page renders and the change is visibly present. If it isn't there, stop — the edit didn't take effect (wrong file, HMR stale, build cache).

2. Interact with the change directly

Reading the DOM is not interacting. Drive it:

  • New control (button / input / toggle / dropdown): click or type into it, then confirm the expected state change actually happened (value updates, panel opens, request fires, list re-renders).
  • Screenshot before and after the interaction. The diff between the two is your evidence.
  • Exercise the states that matter for this change: loading, empty, error, hover/focus, disabled.
  • Check dark mode and mobile width if the change touches layout, color, or spacing.

3. Check the browser console

Open DevTools → Console. Requirement: zero new errors or warnings introduced by your change.

  • React key warnings, hydration mismatches, act() warnings, failed network requests, 404s on assets — all count as failures.
  • Compare against a baseline if the page already logs noise; only new messages are on you.

4. Run a performance pass

If a Chrome DevTools MCP (or equivalent browser-automation tool) is available, use it to record a performance trace and audit Core Web Vitals (LCP, INP, CLS). Prefer whatever project-specific wrapper already drives that tooling over hand-rolling a trace from scratch — but if none exists, run the MCP/Lighthouse directly rather than skipping the step.

  • Flag regressions: layout shift from the new element, a slow interaction (INP), a render-blocking asset you introduced.
  • Scale the rigor to the change: a one-line copy fix doesn't need a full trace; a new above-the-fold component or a data-heavy view does.

The Rule

If any step fails, fix the issue and rerun from step 1. Do not hand back partially verified work, and do not report the earlier steps as passing when a later one is still red — the change is either fully verified or it is not done.

Verification Checklist

State this back before declaring done:

CheckPass?
Dev server running; edited route opened in a browser
Change is visibly present on the page
Interacted with the control; expected state change confirmed
Before/after screenshots captured
Relevant states exercised (loading/empty/error, dark mode, mobile)
Console: zero new errors or warnings
Perf trace / Core Web Vitals: no regression (scaled to change size)

If any row is blank, the task is not done.

Anti-Patterns

PatternProblemInstead
"Edit succeeded, so it works"Compiling ≠ rendering correctlyOpen the page and look
Reading the DOM / component state instead of clickingState can be right while the handler is unwiredActually click it
Screenshotting only the "after"No proof the change did anythingBefore and after
Ignoring console warnings ("just warnings")React key/hydration warnings are real bugsZero new console output
Verifying on the home page, not the edited routeYou never saw your changeOpen the exact path you touched
Reporting steps 1–3 green while step 4 is unrunPartial verification reads as completeNot done until all steps pass

Gotchas

  • HMR can lie. Hot-reload sometimes shows a stale render or a half-applied change. If something looks off, hard-refresh (or restart the dev server) before concluding your code is wrong.
  • Open the right port/route. In multi-service projects, each service can run on its own fixed port. Verifying on the wrong port silently shows an old build from a different service — confirm the URL actually maps to the project you edited.
  • Content baked at boot. Some projects compile content (Markdown, CMS data) into a database or bundle at server startup rather than serving it live. A content edit won't appear until that process restarts — a blank or stale page here is a restart problem, not a frontend bug.
  • Dark mode and mobile are separate surfaces. A change that looks right in light/desktop can break contrast or overflow in dark/mobile — check both whenever the change touches color, spacing, or layout.
  • Scrollable tables and anchor offsets are easy to skip. If the project has house rules for these (e.g. wide tables get overflow-x: auto instead of blowing out page width, anchor links use scroll-margin-top on the heading so it lands visible under a sticky nav), verify they still hold after your change.
  • Chrome DevTools MCP may be absent in headless/cron/CI runs. If the perf tooling isn't reachable, say so explicitly and fall back to a manual Lighthouse or visual check — don't silently skip step 4 and imply it passed.
  • Screenshots are the deliverable, not decoration. When you report the change as verified, attach or reference the before/after so the claim is auditable — "I verified it" without artifacts is the thing this skill exists to prevent.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.