Repository navigation
cc-status shim paints subagent seats in their own colours - #97
Merged
Merged
Conversation
A tab now says which seat is working: plan-fable, unstick-fable, review-fable, work-sonnet and scout-haiku each get a dot and text colour (tabstatus.color.<agent_type> overrides one or adds a seat), while the status stays one of cc-status's three words, which the Session Status tool sorts by. Hooks fired inside a subagent carry agent_type. An in-seat PreToolUse/PostToolUse is painted by the shim instead of cc-status's plain orange; PermissionRequest and Notification pass through, since waiting outranks a colour. SubagentStart and SubagentStop keep a per-session state file under $XDG_STATE_HOME/tabstatus, deleted on SessionStart/SessionEnd and ignored past a day, so a Stop over a still-running background seat repaints its colour instead of orange. Every detail the shim writes now starts with the repo, the basename of the payload's cwd. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
`it2 set-status` leaves fields it is not given unchanged, so a bare `tabstatus idle` over a stuck orange row left the dot orange, and an idle that did not store a background-task count of 0 was revived to "working" by cc-status's idle_prompt handler a minute after Stop. `tabstatus idle|working|waiting` now appends cc-status's dot/text colour pair (unless a colour flag is given) and, for idle, `--background-tasks 0` (unless a count is given), making it the manual reset for a stuck row. Free-text status words and the no-argument clear are unchanged. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
claude-doctor gains a tabstatus: line counting, from a profile's settings.json, how many of the eleven hook events run the shim, how many still run iTerm2's raw cc-status and how many neither, with the shim's deploy path and the live seat lines. iterm2-doctor gains a cc-status: line reading the app's symlink and where it2 is, by readlink and -x only. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Rewrite the CLAUDE.md Session Status passage as one account: quiet vs active by output-file mtime, why the stored count is 0, amber not blue, the full field set, seat colours, the eleven wired events and what Reinstall/Uninstall do to them, the 3.7.3 expanded-path rule, cost, what the shim cannot see, and cc-status explain as the diagnostic. Add "Session Status wiring" to the private-overlays guide: the stat-gated template rule with chezmoi.homeDir, adding SubagentStart, and claude-doctor's tabstatus line as the check. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
hadees
force-pushed
the
feat/cc-status-seats
branch
from
October 6, 2026 19:53
760acad to
4cde696
Compare
…iting A session that ends its turn by asking the operator something in prose showed as idle, with only the message's opening as the detail: cc-status reserves "waiting" for PermissionRequest and AskUserQuestion and paints a plain Stop idle. On Stop and StopFailure with nothing active, the shim now takes the last non-empty line of last_assistant_message, strips leading bullets and quotes, and when it ends in ? or U+FF1F paints waiting in yellow (tabstatus.color.question, default #ffd75f) with that line as the detail. The paint follows cc-status's and outranks the quiet-task idle; SubagentStop never fires it. The sh gate now hands such Stops to python, so python runs on most Stops; explain prints a question= line. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
A backgrounded installer sat 75 minutes on one download at 0% CPU and the row still said working; quiet alone cannot tell that from a long compile that prints nothing. For each quiet shell task the shim now finds the holders of its output file with lsof -t and sums the CPU of their process trees from one ps, keeping the last reading per task beside the seats file. A task nothing holds, or one whose CPU and file have not moved for tabstatus.stall-minutes, is stalled and paints waiting in red; the quiet paint gains its age and a (cpu busy) mark. No lsof means unknown, never stalled. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #96 — fix-plan steps 2–5.
agent_type, so an in-seat PreToolUse/PostToolUse is painted by the shim in that seat's colour instead of cc-status's orange;SubagentStart/SubagentStopkeep a one-line-per-seat state file under${XDG_STATE_HOME:-~/.local/state}/tabstatus/<session_id>.seats, so a parent's Stop over a still-running background seat shows the seat's colour with the active count. Defaults: plan-fable#d787ff, unstick-fable#ff5faf, review-fable#af87ff, work-sonnet#00afaf, scout-haiku#87af5f; override or add any agent type withgit config tabstatus.color.<agent_type> '#rrggbb'. The repo name (basename ofcwd) prefixes every detail the shim writes. The three status words are unchanged, so the panel's sorting and watches keep working.tabstatusfull field set (step 3).tabstatus idle|working|waitingnow sends cc-status's colour pair and, for idle, a zero count — the one-command manual reset for a stuck row (it2 set-statuskeeps every field it is not given).claude-doctor:tabstatus: shim on <n>/11 events, raw cc-status on <m>, none on <k> in <settings>; shim: <path|MISSING>; seats live: <n>.iterm2-doctor:cc-status: ~/.config/iterm2/cc-status -> <target> (<state>); it2: <where>. Nothing executed, readlink and -x only.?) paints waiting in yellow (tabstatus.color.question, default#ffd75f) with the question line as the detail — cc-status alone paints that turn idle with the message's opening, so a session asking you something looked finished. Never on SubagentStop; nothing when work is still active.lsoffinds the operator-owned shell holding its output file, onepssums the CPU time of that process tree (children may be root's;psshows them anyway), and a task whose file and CPU have both stood still fortabstatus.stall-minutes(default 15), or whose holder is gone, paints waiting in red (tabstatus.color.stalled, default#ff5f5f) as "stalled N min: …". The quiet paint now says how long ("quiet M min") and "(cpu busy)" when a silent task is still computing. Live-checked against a download that sat 88 minutes at 0% CPU today.docs/private-overlays.mdgains "Session Status wiring" describing the stat-gated template rule the overlays use and the eleventh event (SubagentStart) to add.Unverified until the first live run: that
SubagentStart/SubagentStopand in-seat tool hooks carryagent_typein the compact shape the sh gate matches (documented, not yet captured); a mismatch falls open to today's behaviour.Tests: full suite 618/618 locally (bash and dash for the shim file).
🤖 Generated with Claude Code