Ballmac UI home

Agent 5: eval results

Evaluation results for a prompt or model change: a verdict that calls out regressions, each metric against the baseline with which way is better, cases filtered to regressions or improvements, and both answers side by side or as a diff.

State

The normal view. Pick a state to see what the block shows while it loads, is empty or fails.

Open in v0 (opens in a new tab)

Installation

$ pnpm dlx shadcn@latest add @ballmac/agent-5

Usage

Example: agent-5-demo.tsx
"use client"

import { Agent5 } from "@/components/ballmac/blocks/agent-5/agent-5"

export default function Agent5Demo() {
  return <Agent5 onShip={() => {}} onKeepBaseline={() => {}} />
}

States

Pass state and the block does the rest: skeletons that match its layout, a clear empty message, an error you can retry, and announcements for screen readers.

  • Loading

    Skeleton metric cards and a case list under the real heading.

    state="loading"
  • No results yet

    Asks to run the suite; emptyState can add a Run button.

    state="empty" (or cases={[]})
  • Could not load

    An alert with Try again (onRetry).

    state="error"

Quality scorecard

8/8

Every applicable check passes

Measured in Chromium 153.0.8010.12 on October 19, 2026, on version 1.0.0, across 4 examples.

How we measure
  • Passed: Accessibility scan

    No serious or critical issues in light and dark, on a phone and a desktop (axe-core 4.13.0, WCAG 2.1 AA and best-practice rules).

  • Passed: Small screens

    No sideways scrolling at 360, 390 or 1280 px wide.

  • Passed: Reduced motion

    Every endless animation stops when the visitor asks for less motion.

  • Passed: Right to left

    Mirrors correctly in Arabic, Hebrew, Persian and Urdu layouts (8% of elements off the mirror, limit 15%).

  • Passed: Clean console

    No console errors or React hydration errors while loading.

  • Passed: Keyboard

    10 tab stops, each showing a visible change when it takes focus. 3 documented keys.

  • Passed: Loading, empty and error

    Ships loading, empty, error states, each with an example.

  • Passed: Automated tests

    Covered by agent-5.test.tsx.

  • For information: Footprint

    4.8 KB of source (gzip) in 1 file. Installs lucide-react.

Automated checks find a subset of problems and do not replace testing with a screen reader and a keyboard in your own app. What they miss.

Dependencies

Use with AI

An evals comparison page. <Agent5 title baselineLabel candidateLabel metrics={[{ id, label, baseline, candidate, format: 'percent' | 'ms' | 'currency' | 'number', higherIsBetter }]} cases={[{ id, input, expected, baseline: { output, pass, score }, candidate: { output, pass, score }, tags }]} onShip onKeepBaseline />. A regression is a case that passed on the baseline and fails on the candidate; the first one is selected. With the shadcn MCP server set up (guide), ask your agent:

Add the Ballmac UI Agent 5: eval results (@ballmac/agent-5) to this project with the shadcn MCP, then use it where it fits.

Use it for

  • Deciding whether a prompt or model change can ship
  • Reviewing an eval suite's results in an internal AI tool

Not for

  • Trying prompts by hand (use agent-4)
  • Tracing one run (use agent-3)

Registry JSON: https://ui.ballmac.com/r/agent-5.json

Free to use in personal and commercial projects.

Ballmac UI Pro

Need the whole screen, not one piece?

150 premium blocks (dashboards, settings, calendars, mail, shops, onboarding), a private shadcn registry and Figma tokens. Blocks that show data come with loading, empty and error states built in.

Not ready? Try three Pro blocks free

Leave your email and the blocks appear right here, ready to copy. Yours to use in any project.