Skip to content

No.149 · AI & Chat

Agent Run

A coding agent's work as a checklist that fills in while it happens — search, read, edit with a diff stat, run tests — where any finished step opens to show its output.

The prompt

The design brief behind this component — reuse it to regenerate, remix, or explain it.

Present an AI agent's run as a compact dark panel (#111, 14px radius, hairline ring). The header has a small agent mark, the task in one line, and a mono status ('Step 3 of 6' → green 'Done · 7.2s'). Each tool call is a row: a verb glyph (think / search / read / edit / run, 16px line icons), a muted verb ('Searched for', 'Read', 'Edited', 'Ran'), the target in a mono pill ('src/lib/pricing.ts', 'npm test -- pricing') and a small meta ('+6 −1' with a green first character, '12 passed'), and a status slot on the right. Queued rows sit at 35% opacity; the active row brightens, its glyph turns accent and a spinner turns; completed rows get a tick that draws itself. Rows with output (a diff, a test log) become buttons once complete and fold open (grid rows 0fr → 1fr, 420ms) to a small mono block with coloured +/− lines. When the last step lands, a one-paragraph summary folds open beneath with 'Run again'.

Notes

Interaction
Watch the run progress; click a finished Edit or Run row to see its diff or output; 'Run again' replays.
Motion
Active row rises in 420ms; spinner → tick (360ms draw); details and summary fold open 420–520ms.
Accessibility
Semantic section with an ordered list; expandable rows are buttons with aria-expanded; completion is announced politely. Reduced motion shows the finished run immediately.
Responsive
Fluid up to 36rem; long targets truncate with an ellipsis.
On touch
Tap targets are full-width rows.

Featured

301 components