subagentStatusLine only covers Task subagents, not teammate or workflow rows in the same panel

What happens

subagentStatusLine only receives Task-tool subagents. Teammate rows and workflow rows sit in the same agent panel, directly above and below those subagents, but never reach the command — so they keep the built-in format while the subagent rows use mine. Neighbouring rows in one panel end up looking like they come from two different programs.

Version: Claude Code 2.1.218, macOS.

How I verified it

My subagentStatusLine command logs the JSON it receives on stdin before rendering.

Task subagents — works. The tasks array arrives as documented, every entry "type": "local_agent", model resolved (claude-sonnet-5, claude-opus-4-8[1m]), and my rows render.

Teammates — never delivered. With CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1, teammate rows (@counter, @reader) rendered in the default format while Task subagents in the _same panel, at the same moment_ rendered mine. No teammate entry ever appeared in the logged tasks array.

Workflows — never delivered. Two control runs (13 s and 24 s; the second with "refreshInterval": 2 to rule out an idle panel). The workflow row was visible in the panel for the whole time. The command was not invoked once — the log file was never created.

So the panel row for a workflow stays as name · description · 0/12 agents done · 2m 49s · ↓ 866.0k tokens with no way to reformat it, and there is no model information on teammate rows at all.

What I'd like

subagentStatusLine to cover every row the agent panel shows — Task subagents, teammates, and workflows — with the existing per-task fields where they apply (model, effort, contextWindowSize, tokenCount) plus whatever a workflow row needs for its aggregate progress (e.g. agents done / total). The documented type field already distinguishes them, so a script can format each kind differently; today it only ever sees one kind.

If covering all three is not intended, it would help for the docs to say plainly which row kinds the setting applies to. The current wording — "each subagent shown in the agent panel below the prompt" — reads as if it covers the whole panel, which is what led me to spend the debugging time above.

Related

The main statusLine always reports the main session's model, even while a subagent transcript is focused in the TUI. subagentStatusLine solves this for panel rows, so the remaining gap is the bottom bar: there is no field for the currently focused agent's resolved model. Closest existing issues I found are #76370 and #24094, both about the progress/task UI rather than the status bar.

Docs

https://code.claude.com/docs/en/statusline#subagent-status-lines