Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build or adapt a local browser/CDP harness to drive and inspect a web, IDE, or Electron UI. Use for local UI verification, screenshots, accessibility snapshots, perf profiles, visual diffs, or reproducing UI bugs.
.claude/skills/asymmetric-al-control-ui/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-07 | ✓→✓ | = Same ✓ | 23% | 0% |
| case-08 | ✓→✓ | = Same ✓ | -7% | 0% |
Use local browser automation to verify UI behavior with evidence. First reuse the repo's own Playwright, browser, or Electron harness if it exists; otherwise assemble a temporary local harness around the app's dev server or Chromium debug port.
verify-this.data-* selectors over coordinates.Use the repo's installed browser tooling when possible. If the repo already has Playwright, a minimal one-off probe looks like:
javascriptimport { chromium } from "playwright"; const browser = await chromium.launch(); const page = await browser.newPage({ viewport: { width: 1280, height: 800 } }); await page.goto("http://127.0.0.1:<port>"); await page.getByRole("button", { name: /submit/i }).click(); await page.screenshot({ path: "/tmp/ui-harness-after.png", fullPage: true }); await browser.close();
Do not add Playwright as a project dependency just for this probe unless the user asks. Prefer existing dev dependencies or external browser tools already available in the environment.
For Electron or a Chromium app launched with --remote-debugging-port=<port>, connect over CDP:
javascriptimport { chromium } from "playwright"; const browser = await chromium.connectOverCDP("http://127.0.0.1:<debug-port>"); const pages = browser.contexts().flatMap((context) => context.pages()); let page; for (const candidate of pages) { if (await candidate.locator("<app-root-selector>").count()) { page = candidate; break; } } if (!page) { console.log( await Promise.all( pages.map(async (p) => ({ title: await p.title(), url: p.url(), })), ), ); throw new Error("No matching app page found"); } await page.screenshot({ path: "/tmp/ui-harness-cdp.png", fullPage: true }); await browser.close();
Replace <app-root-selector> with a stable marker from the current repo, such as a root app node, landmark, or product-specific data-* attribute.
Use raw CDP only when higher-level browser APIs are insufficient:
When multiple app windows/tabs share a debug port:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 8,349 | 5,373 | -36% | 1 | 1 | 0% | 1,528 | 1,385 | -9% | 0 | 0 | — |
case-07 | pass→pass | 18,490 | 14,578 | -21% | 1 | 1 | 0% | 2,821 | 3,470 | +23% | 0 | 0 | — |
case-02 | fail→fail | 10,237 | 6,245 | -39% | 1 | 1 | 0% | 1,817 | 1,326 | -27% | 0 | 0 | — |
case-03 | fail→fail | 8,426 | 4,401 | -48% | 1 | 1 | 0% | 1,460 | 1,240 | -15% | 0 | 0 | — |
case-04 | fail→fail | 14,661 | 9,263 | -37% | 1 | 1 | 0% | 2,206 | 2,595 | +18% | 0 | 0 | — |
case-05 | fail→fail | 13,251 | 12,342 | -7% | 1 | 1 | 0% | 2,079 | 3,011 | +45% | 0 | 0 | — |
case-06 | fail→pass | 12,769 | 6,963 | -45% | 1 | 1 | 0% | 1,904 | 2,148 | +13% | 0 | 0 | — |
case-08 | pass→pass | 13,209 | 5,302 | -60% | 1 | 1 | 0% | 1,936 | 1,791 | -7% | 0 | 0 | — |
case-09 | fail→fail | 11,805 | 8,026 | -32% | 1 | 1 | 0% | 1,916 | 2,380 | +24% | 0 | 0 | — |
case-10 | pass→pass | 12,426 | 5,732 | -54% | 1 | 1 | 0% | 1,919 | 2,047 | +7% | 0 | 0 | — |
case-11 | pass→pass | 15,054 | 5,346 | -64% | 1 | 1 | 0% | 2,422 | 1,855 | -23% | 0 | 0 | — |
case-12 | pass→pass | 15,494 | 12,689 | -18% | 1 | 1 | 0% | 2,425 | 3,149 | +30% | 0 | 0 | — |
case-17 | fail→pass | 12,320 | 5,861 | -52% | 1 | 1 | 0% | 2,146 | 2,143 | -0% | 0 | 0 | — |
case-13 | pass→pass | 9,207 | 3,951 | -57% | 1 | 1 | 0% | 1,380 | 1,689 | +22% | 0 | 0 | — |
case-14 | pass→pass | 10,103 | 9,677 | -4% | 1 | 1 | 0% | 1,602 | 2,753 | +72% | 0 | 0 | — |
case-15 | pass→pass | 14,453 | 10,563 | -27% | 1 | 1 | 0% | 2,091 | 2,612 | +25% | 0 | 0 | — |
case-16 | fail→pass | 14,418 | 4,506 | -69% | 1 | 1 | 0% | 2,460 | 1,866 | -24% | 0 | 0 | — |
case-18 | pass→pass | 11,436 | 10,567 | -8% | 1 | 1 | 0% | 2,175 | 3,136 | +44% | 0 | 0 | — |
case-19 | pass→pass | 13,571 | 12,301 | -9% | 1 | 1 | 0% | 3,072 | 3,829 | +25% | 0 | 0 | — |
case-20 | pass→pass | 13,037 | 8,040 | -38% | 1 | 1 | 0% | 2,961 | 2,844 | -4% | 0 | 0 | — |
case-21 | pass→pass | 14,795 | 15,414 | +4% | 1 | 1 | 0% | 3,035 | 4,300 | +42% | 0 | 0 | — |
case-22 | pass→pass | 6,933 | 4,121 | -41% | 1 | 1 | 0% | 1,324 | 1,841 | +39% | 0 | 0 | — |
case-23 | pass→pass | 10,031 | 3,901 | -61% | 1 | 1 | 0% | 1,886 | 1,817 | -4% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 20 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +13 percentage points is the difference between those two pass rates over the 20 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.