Direct answer: a reviewable AI coding agent output should contain the request, acceptance criteria, failing tests, scoped implementation notes, and PR evidence a human can verify.
Acceptance criteria
- Table uses the existing DataTable, Toolbar, Avatar, and StatusChip components.
- Rows stay readable at desktop, tablet, and mobile widths.
- Long owner names truncate without pushing action buttons off-screen.
- Empty, loading, and error states match the product surface.
Tests before code
- Component test for row rendering with long names.
- Interaction test for toolbar action availability.
- Accessibility check for status labels and keyboard focus order.
- Visual review screenshots attached to the PR.
Implementation notes
- Map screenshot regions to existing component primitives before code changes.
- Keep spacing and type scale inside current token values.
- Avoid creating one-off CSS unless an existing primitive cannot express the layout.
PR review evidence
- PR includes before/after screenshots.
- Reviewer can compare responsive states against the visual spec.
- Ambiguities are listed instead of silently guessed.