[jamesyong-42]

What

A host input, waitMark (R10, RoomWaitMark), that lets a host hand the room its own mark for the room's
waits. With one, the room draws its waits in that mark. With none, every wait stays as rosetta's app shows it today.
The reference host passes none, so the app reads byte for byte as before.

• mark(place, { resting }) draws the mark for one place, "view", "column" or "card". The mark is at rest
  while the work it stands for is paused.
• carried says the host's own hold already stands where the room's first one comes. The room's first hold
  then stands at once, so the mark is never faded out and shown again.

What changes when a mark is given:

• The view's waits are one hold. Reaching the index, then the cold start before its first project
  (HistoryRoom.tsx, components/ui/Hold.tsx):
  • the mark stands at the centre of the room's box;
  • one line 26 under it: connectingTitle, or the step's own title with the ellipsis of work under way
    ("Finding your history…", "Discovering conversations…");
  • its amount 4 under that ("1,333 files checked", a paused import's "80 sessions found"), in tabular figures. The
    first step's title stands alone, because its amount ("Looking for conversations…") counts nothing.
• The hold's timing (lib/use-standing-hold.ts):
  • It is late, as the veil is: nothing for 120ms.
  • Once it has shown it stands at least 800ms, then fades over 280ms (120ms under reduced motion).
  • What comes next waits for it.
  • A new line crossfades in place, 140ms out and 240ms in.
• A failure is no wait. An index that has stopped answering keeps the room's own face.
• The import card wears the mark where IMPORTING HISTORY stood, with the step beside it (IngestStatus.tsx):
  • its words keep to the step's title and its count;
  • a fraction the step knows is a colourless 2px hairline, and a step that knows none draws no bar;
  • the detail, the amount, the recency line, the connection and the timer fold under Agent details, which now
    stands through search too;
  • paused updates stand the mark at rest.
  • The card without a mark is unchanged. Its live regions, footer and agent list are now parts that both cards
    draw.
• A conversation's wait is a hold in its column (TranscriptLoading.tsx), its sketch and sweep gone. "Read
  available messages" stays under its words. A load that failed keeps its own face and Try again.
• Nothing sweeps beside the mark (Meter.tsx). A meter that knows no amount draws nothing, a veil keeps its
  words, and a meter that knows still fills.
• PAUSED_UPDATES holds the card's paused words in one place (lib/ingest-activity.ts). The card and the hold both
  read them, so they cannot drift.

Why

VibeField's History mounts this room. Its loads elsewhere (the Setup Assistant's) are one look: the field's living
mark and one line under it. Today History's waits are each in their own look:

• the connecting face's sweep;
• the cold start's centred card;
• the import card's eyebrow and its --vf-select bar;
• the transcript's sketch and sweep;
• the veils' sweeps.

The card's bar is selection blue, against the room's own word for its meter ("The room's colourless meter",
Meter.tsx).

This keeps the room's waits the room's own: their steps, their words, their place ("Progress describes its named
step and never covers available content"). It takes only the mark a host places in them, as the room already takes
connectingTitle and the consent slot. VibeField's ruling: track HG, HG-W1…W5 and HG-L1, DESIGN.md §8 "History".

Asks (the owner's to rule)

• A1, one input with places, mark(place, { resting }), rather than three node props. The room asks for the
  mark where it waits, and the host draws it at the size the place needs (VibeField: 99 wide in a hold, 40 on the
  card).
• A2, a mark means a hold. With a mark, the connecting face and the cold start become the hold. Without one, the
  room's faces are untouched. Is that coupling right, or should the hold be its own option?

Tests

clients/app/test/room-wait-mark.test.mjs, 9 tests, server renders and the timing's pure function:

• with no mark, the room keeps its own waits;
• in a mark, the connecting face is a hold: the host's title and words, no sweep, no face;
• carried: no late;
• the floor: gone before it shows, standing 800ms, then leaving;
• the cold start's words:
  • "Finding your history…" alone;
  • "Discovering conversations…" over its count;
  • a paused import's title and count at rest;
  • paused updates' words at rest;
• the card: the mark, the step, the title and its count, the hairline only for a fraction, the rest under Agent
  details, at rest while paused;
• meters and veils beside the mark;
• the column hold, its act, and a failed load's own face.

Run on the author's Mac, one file at a time, on Node 26.5.0 and npm 12.0.1 after npm ci --ignore-scripts:

• the protocol's and the room's builds (tools/npm/build.mjs). The room's build is its typecheck, with
  declarations.
• room-wait-mark 9/9, dom-contract 15/15, import-state 21/21, ingest-progress 11/11, room-meters 5/5,
  room-not-answering 3/3, room-package 4/4 and room-search-key 4/4. These are every app test that imports a module
  this changes, or walks the room's sources.

Proved by rosetta CI at this PR, not run here:

• the rest of npm test and the app's build;
• the docs check and check:packages;
• the Electron fixtures (room-host, cold-import).

No server render drives the crossfade's effect, and no fixture covers it.

Out of scope

• The reader's (TranscriptReader) own opening face: VibeField's Peek keeps it as it is.
• The rows' and the charts' waits: they keep their own ink and words.
• The release: this rides the next one, 0.1.6.

🤖 Generated with Claude Code

---

^{Open in [code]smith · Need help on this PR? Tag @codesmith-bot with what you need.}