tui-tab-bar — Workgraph live mirror

Metadata

Status	done
Assigned	`agent-715`
Agent identity	`3184716484e6f0ea08bb13539daf07686ee79d440505f1fdf2de0357707034c3`
Created	2026-04-27T15:43:59.295676290+00:00
Started	2026-04-27T15:45:55.870348589+00:00
Completed	2026-04-27T15:59:40.811468031+00:00
Tags	`eval-scheduled`

Description

User report: at the top of the TUI screen, chat tabs render as:

[1] ● coord:1 ✕ │ [2] ● coord:3 ✕ │ [3] ◉ coord:4 ✕ [+]

Two problems:

Deprecated terminology — labels say coord:N, but per CLAUDE.md the words 'coordinator' / 'orchestrator' are deprecated as role-nouns. Chat task ids are .chat-N (e.g. .chat-2, .chat-3, .chat-4 — visible in the graph viz column on the left of the user's screenshot). Tab labels should match the task id: .chat-2, .chat-3, .chat-4.
Wrong numbering — first tab shows coord:1 but no .chat-1 exists in the graph; user has .chat-2, .chat-3, .chat-4. Looks like the tab bar is using a separate stale 1-indexed counter instead of reading the actual chat task id. User said: 'what is 1??? why doesn't it say .chat-4 etc.'

The leading [1]/[2]/[3] is the user-facing tab number (fine to keep as positional shortcut for hotkeys). The label after the dot/circle indicator is what needs to come from the chat task id.

Files likely to touch

src/tui/ — wherever the chat tab bar is rendered. Look for tab-label builders that format 'coord:' or use a tab-index-derived integer instead of the chat ref / task id.
The chat ref scheme (per memory feedback_chat_ref_vs_task_id): .chat-N is the task id, chat-N (no leading dot) is the chat alias. Tab labels should show the task id form (with leading dot) since that's what wg show / wg list use.

Validation

Failing test first: render the tab bar with a fixture chat list of [.chat-2, .chat-3, .chat-4] → assert label strings contain '.chat-2', '.chat-3', '.chat-4' and contain neither 'coord:1' nor any 'coord:' prefix
Implementation makes test pass — read tab label from chat task id, not from a positional counter or hardcoded 'coord:' prefix
cargo build + cargo test pass with no regressions
Manual smoke: open wg tui with multiple chats; tab labels match wg list --status open task ids exactly
Grep src/tui/ for any remaining 'coord:' literal in tab/label rendering — none should remain

## Description

User report: at the top of the TUI screen, chat tabs render as:

```
[1] ● coord:1 ✕ │ [2] ● coord:3 ✕ │ [3] ◉ coord:4 ✕ [+]
```

Two problems:

1. **Deprecated terminology** — labels say `coord:N`, but per CLAUDE.md the words 'coordinator' / 'orchestrator' are deprecated as role-nouns. Chat task ids are `.chat-N` (e.g. `.chat-2`, `.chat-3`, `.chat-4` — visible in the graph viz column on the left of the user's screenshot). Tab labels should match the task id: `.chat-2`, `.chat-3`, `.chat-4`.

2. **Wrong numbering** — first tab shows `coord:1` but no `.chat-1` exists in the graph; user has `.chat-2`, `.chat-3`, `.chat-4`. Looks like the tab bar is using a separate stale 1-indexed counter instead of reading the actual chat task id. User said: 'what is 1??? why doesn't it say .chat-4 etc.'

The leading `[1]`/`[2]`/`[3]` is the user-facing tab number (fine to keep as positional shortcut for hotkeys). The label after the dot/circle indicator is what needs to come from the chat task id.

## Files likely to touch

- src/tui/ — wherever the chat tab bar is rendered. Look for tab-label builders that format 'coord:' or use a tab-index-derived integer instead of the chat ref / task id.
- The chat ref scheme (per memory feedback_chat_ref_vs_task_id): `.chat-N` is the task id, `chat-N` (no leading dot) is the chat alias. Tab labels should show the **task id** form (with leading dot) since that's what `wg show` / `wg list` use.

## Validation

- [ ] Failing test first: render the tab bar with a fixture chat list of [.chat-2, .chat-3, .chat-4] → assert label strings contain '.chat-2', '.chat-3', '.chat-4' and contain neither 'coord:1' nor any 'coord:' prefix
- [ ] Implementation makes test pass — read tab label from chat task id, not from a positional counter or hardcoded 'coord:' prefix
- [ ] cargo build + cargo test pass with no regressions
- [ ] Manual smoke: open wg tui with multiple chats; tab labels match `wg list --status open` task ids exactly
- [ ] Grep src/tui/ for any remaining 'coord:' literal in tab/label rendering — none should remain

Depends on

done .assign-tui-tab-bar

Required by

(none)

Log

2026-04-27T15:43:59.289685339+00:00 Task paused
2026-04-27T15:44:28.895707599+00:00 Task published
2026-04-27T15:45:55.870352236+00:00 Spawned by coordinator --executor claude --model opus
2026-04-27T15:46:09.083300808+00:00 Starting work — checking TUI tab bar rendering code
2026-04-27T15:54:27.691536669+00:00 Failing test added (passed cid in [.chat-2, .chat-3, .chat-4] fixture). Fix: list_coordinator_ids_and_labels now uses format_chat_task_id() instead of 'coord:N'. Smoke scenario tui_chat_switch_keystrokes.sh updated to grep .chat-N + reject any coord:N regression.
2026-04-27T15:59:19.814612222+00:00 Committed: 78099f3bf — pushed to remote
2026-04-27T15:59:40.811482317+00:00 Task pending eval (agent reported done; awaiting `.evaluate-*` to score)
2026-04-27T16:01:08.681878949+00:00 PendingEval → Done (evaluator passed; downstream unblocks)