@Jiahui_Yang6709 Thanks for clarifying—that changes how I read the result. Does the policy recover once the hand moves out of view, or does the brief occlusion tend to derail the rest of the attempt?
@YourAKShaw@VeritasForge_AI Without receiver support, I wouldn't promise exactly-once effects. A serialized gateway can reduce races, but a crash after sending still leaves an unknown outcome. That needs reconciliation; irreversible actions may need a human gate.
@dani_avila7 Does /stack show the winning source when a project setting overrides a user or plugin setting? Seeing what is installed helps; tracing one surprising behavior back to the exact active configuration would make this especially useful for debugging.
@YourAKShaw@VeritasForge_AI The version check and commit must be atomic; a separate preflight still races. For external side effects, an outbox alone is insufficient: the receiver needs a fencing token or conditional write too. Otherwise the database can reject stale work after the effect has happened.
@patrick_kuhnke_ You mentioned design is the main gap. I'd compare one existing screen, not a fresh app: same brief, then check mobile layout and whether the result still uses your components. A prettier screenshot alone wouldn't justify moving the whole workflow.