Omit Actions the User Can't Take — Don't Render Them Disabled
The convention is to render every action and disable the ones unavailable. For permission-gated operations, that's the wrong default.
The problem
A disabled "Strike Out" button tells a clerk that striking out exists, that it applies to this record, and that someone in the building can do it. Multiply by every action on every screen and your UI has documented your authorisation model for anyone who reads it.
There's a worse version: the button is disabled client-side only, and the endpoint isn't checked. Now it's not just disclosure.
Two different answers
Be explicit that these are different questions:
- Not permitted for this actor → omit the action entirely.
- Permitted, but not possible yet → show it, disabled, with the reason.
public async Task<IReadOnlyList<AvailableAction>> AvailableActionsAsync(
Guid recordId, Actor actor)
{
var transitions = await ResolveTransitionsAsync(recordId);
return transitions
// Silently gone if the actor may not act.
.Where(t => actor.Permissions.Contains(t.RequiredPermission))
.Select(t => new AvailableAction(
t.ActionKey,
Enabled: GuardsPass(t, context),
Reason: FirstUnmetRequirement(t, context)))
.ToList();
}Why it matters
The screen renders from this list, so the UI can't drift from the rules — there's no second copy of the permission logic in a component.
And the distinction is honest. "You can't see this" and "you can't do this yet" are genuinely different messages. Collapsing them into one grey button communicates neither.