Tasks and rounds

Task dashboard

Clicking a task opens a dashboard, not a transcript. It combines:

• Overview — status, assignee, engine, environment, plan, and latest receipt.
• Conversation — user prompts, cited answers, round markers, run summaries, and a task composer.
• Changes — exact files, patches, frames, design edits, and verification evidence.
• Activity — the attributed event history and full-fidelity engine feed.

Code names

Each task and mission receives a code name, such as Meita, alongside its descriptive title. The code name identifies that work’s context and is never reused within the project. It is not an agent or an engine. Naming does not by itself imply that every name-addressed delegation workflow is available. See Work-context files.

Round contract

A round begins when a prompt is armed against the task and ends at a receipt. A steer redirects a live run and stays inside the current round. Editing the task spec or attachments enriches the next round; it is not a round by itself. Engine-native resume is an optimization, never the carrier of task memory.

Run and Ask

The task composer has two explicit actions:

• Run steers a supported live engine or starts round N+1 with the task history folded in.
• Ask answers from project and task context with citations. It creates no round and does not change status.

Continue or split

A same-objective follow-up continues the task. Scope-expanding work becomes a new task in the same mission. When General or Chat re-triage detects both possibilities, it offers explicit continuation chips and never auto-submits the receiving task.