Agent Status Card

v1.0.0·Added in 1.0.0

An AI agent at work at a glance: orb, name, status, current step, progress, elapsed time, steps, logs and the controls that matter now.

The card shows the agent's orb (it thinks while the agent runs, listens while it waits for a person and rests otherwise; it only animates near the viewport and holds still under reduced motion), its name and task, and a status badge whose dot pings while it runs and which pops once with a check when a run finishes with every step green. Under it the current step shimmers while it runs, a progress bar glides from step to step (Step 3 of 5) and the elapsed time ticks every second. Only the controls that matter now appear: Pause and Stop while it runs, Resume and Stop while paused, Reject and Approve in an amber box when it waits for approval, Retry in a red box when it failed, View result in a green box when it finished. Stop asks first in a centered dialog (Escape or Keep running cancels). A handler that returns a promise spins its button and disables the others until it settles; a rejection is announced. The Steps and Logs toggles at the bottom open the step timeline and a monospace log that follows new lines until you scroll up inside it, and scroll with the keyboard once focused. The ⋯ button opens the agent's menu, a bottom sheet on phones. Tab reaches every control in order, Enter and Space press them, and each control's name includes the agent's name (Stop Collections agent). Status changes are announced once to screen readers. The compact variant keeps the orb, name, status, current step, progress and one primary action for grids; the row variant is a 64px list line with a progress ring that opens the agent on click, tap or Enter. Below 400px the orb sits above the text, buttons share the width and grow to 44px on touch, and logs cap at 40% of the screen height.