Change Request — Restore engagement-creation affordance (lens-home) — v0.1

Document: cr-restore-engagement-creation-affordance-v0_1 Version: v0.1 Date: 2026-06-28 Type: Phase change request — frontend only (Operator Layer loomworks repo). No engine, no classifier, no create-page change. Consumer: Claude Code (executor). Branch base: feat-companion-surface @ dac9fec (current OL HEAD; carries the mobile Companion surface + add-side work). Governing scoping note: loomworks-restore-engagement-creation-affordance-scoping-note-v0_1. Grounding: loomworks-create-flow-and-old-surface-omissions-inspection-v0_1 (capability is live, only the entry point is gone) + the CR-drafting grounding pass (real Home.tsx, HomeCards.tsx, create-engagement/page.tsx, TopBar.tsx at dac9fec).

Plain-language summary

You can create an engagement today — the engine, the Companion, and the create page all work — but there's no visible way in from the home where you manage your engagements. The old home had a create button; the new home (built to a mockup that omitted it) doesn't. This adds a create button back to the home that opens the create conversation. One control, both layouts (desktop and mobile), nothing new underneath — it points at the create page that already exists.

What's true today (grounded at dac9fec)

• The capability is live end-to-end. Engine create routes mounted; three classifier intents (create_engagement_entry/_active/_commit) with real handlers; the create page /operator/create-engagement renders ChatView mode="create-engagement" and works. (Inspection-confirmed.)
• The create page takes no params. create-engagement/page.tsx auth-guards then renders <ChatView mode="create-engagement" />; ChatView in create mode sends intent_hint: "create_engagement_entry" on the first turn and operates with no projectId. So the affordance target is a bare router.push("/operator/create-engagement") — nothing to thread.
• The home has no create affordance. Home.tsx desktop header (:289-297) is wordmark + operator name (left) · Avatar (right); the lens-tab row (:298-325) is the three tabs, a flex-1 spacer, then HOME.voiceHint as a static string at top-right (:324). No create button anywhere; grep "Create/New/Start engagement" → 0.
• Correction to the scoping note's "try-prompt" lean: the home's top-right "open a project · what needs me" is not interactive — it's a static HOME.voiceHint string, no click behavior. The real tryPromptHint mechanism lives in TopBar.tsx as a mic-button tooltip, and TopBar is the in-engagement surface, not the home. There is no interactive prompt seam on the home to add a create prompt to. The affordance is therefore a new button, not an added prompt. (This simplifies the CR.)
• The navigation idiom to follow. Every home card is <button onClick={() => onOpen(projectId, displayIdentifier)}> (HomeCards.tsx); onOpen (Home.tsx:113-117) does focusCtx.setFocus(...) then router.push("/operator/engagement/<address>"). A create control follows this idiom — a <button> doing router.push("/operator/create-engagement") — minus the focus-set (no engagement exists yet to focus).

Seed alignment

Aligned — restores a capability the seed's success conditions assume (creating an engagement is the operator's first action). "Only show what is available" cuts toward this: creation IS available (live), so it must appear; its absence is the surface failing to show an available capability — the same omission class as upload / navigation / the mobile transcript.

Decisions (settled, grounded)

1. The affordance is a create button (not a try-prompt — no prompt seam exists on the home). Settled by grounding.
2. Navigate, don't open-in-place — router.push("/operator/create-engagement"), reusing the working create page exactly as-is. The in-place option (rendering the create conversation on the home) is a deliberate later refinement, out of scope here.
3. Both layouts — desktop (the lens-tab row's right cluster, where the static hint sits) and mobile (the home's mobile layout needs a create entry too).
4. Follow the card-button idiom — a <button> with the home's existing button styling, router.push on click, no focus-set.

Decisions deferred / out of scope (named, not built)

• Desktop home voice. Grounding found the desktop home has no mic at all — voice on the home is mobile-only. A spoken "create an engagement" already works on mobile (routes through sendConverse({surfaceContext:"home"}), the engine handles the create intent) but desktop has only the static hint. Surfacing desktop home voice is a real adjacent improvement — but it's a different gap (desktop home voice, already listed in build-queue v0.3 open gaps as "desktop home voice — TBD"), not the create-button gap. Out of scope — folding it in would make this two unrelated changes. Named for its own keep-or-drop.
• The other omissions (new-workspace entry, saved-filters link, system-page links, the deleted engagement table) — separate scoping, per the scoping note.

The build — steps

Per-step commits, suite green at each, halt-before-push. Frontend only; no engine, no migration, no create-page change.

Step 0 — inspect before building (no code)

Confirm against feat-companion-surface @ dac9fec:

1. Branch is at dac9fec.
2. Home.tsx lens-tab row structure (:298-325) — the flex-1 spacer + HOME.voiceHint right-cluster placement is as grounded, so the create button has a home in that cluster.
3. The home's mobile layout (:335-431) — where a create control sits there (the mobile home has Personal pinned above a swipe stack + an anchored bottom bar with a mic). Identify the mobile placement (lean: near the header or as a visible control above/within the zone list — NOT hidden in the bottom bar, which is the voice composer).
4. The existing button/tab styling tokens to match (the lens-tab buttons and the CARD class in HomeCards.tsx) so the create button reads as native chrome, not a bolt-on.

Report findings, then build.

Step 1 — Desktop create button

In Home.tsx, add a create button to the lens-tab row's right cluster (:298-325), beside or in place-of-prominence with the static HOME.voiceHint. The button:

• Renders the create label (new HOME.createEngagement string — see Step 3).
• onClick → router.push("/operator/create-engagement"). (The router is already in scope in Home.tsx — it drives onOpen's push.)
• Styling matches the home's existing control vocabulary (lens-tab button / card styling) — a visible, native-looking "act on the whole space" control. Lean: a bordered button consistent with the lens-tab row, distinguished as an action (not a tab).
• Decide with the static hint: keep HOME.voiceHint beside the button, or let the button replace it as the right-cluster element. Lean: keep the button as the primary right-cluster affordance; the static hint can sit beside it or be dropped (it described voice discoverability, which the button doesn't replace — so keep it on mobile where voice is wired; on desktop where there's no mic, the hint is arguably misleading and the button is the real affordance). Confirm at Step 0 against the real layout.

Suite expectation: new visible control; existing home tests unaffected (additive). Add a test: the create button renders on the desktop home and navigates to /operator/create-engagement on click.

Step 2 — Mobile create button

In Home.tsx's mobile layout (:335-431), add the same create affordance, placed per the Step-0 finding (lean: a visible control near the header or above the zone list — not buried in the voice bottom bar). Same router.push("/operator/create-engagement") handler.

Suite expectation: add a test: the create button renders on the mobile home and navigates on click. Existing mobile-home tests unaffected.

Note: Steps 1 and 2 may collapse into one commit if the create button is a shared element both layouts render (cleaner — one button component, placed in each layout). Lean: a single shared CreateEngagementButton (or inline shared element) rendered in both layouts, one commit, rather than two divergent inline buttons. CC's call at build time based on the real layout structure — but prefer the shared element to avoid two buttons drifting.

Step 3 — Strings

Add to HOME in src/lib/strings.ts:

• createEngagement: "Create engagement" — the button label. (Plain-terms; "engagement" is the seed's word, wall-safe.)

Confirm the vocabulary wall accepts it (it uses "engagement" already throughout HOME).

Acceptance

• Create button visible on the home — desktop (lens-tab right cluster) and mobile (per Step-0 placement). It reads as native home chrome.
• Clicking it opens the create conversation — navigates to /operator/create-engagement, which renders the working create flow. End-to-end: button → create page → (Operator converses) → engagement created. (The create flow itself is untouched and already works; this CR's acceptance is that the button reaches it.)
• No regression — the lens-home's zones, lenses, voice path, and navigation are unchanged; the button is purely additive.
• Suite green at each step; OL typecheck + lint + vocabulary-wall clean. No engine change, no migration.
• Provenance/grant conformance (build-queue cross-cutting): N/A for a navigation button — it creates nothing itself; the create flow it reaches carries its own provenance. Named for completeness.

What this does NOT do (scope boundary)

• No engine, classifier, or create-page change — all live and untouched.
• No open-in-place create conversation on the home (navigate-to-page only; in-place is a later refinement).
• No desktop home voice (separate gap, named).
• No other omissions (new-workspace, system pages, etc. — separate scoping).
• No merge to main — lands on feat-companion-surface; merge is a separate authorization.

DUNIN7 — Done In Seven LLC — Miami, Florida Change Request — Restore engagement-creation affordance (lens-home) — v0.1 — 2026-06-28 Adds a create-engagement button to the lens-home (desktop + mobile) doing router.push("/operator/create-engagement"), following the card-button idiom minus focus-set. The capability is live end-to-end; only the entry point was dropped to a mockup. Frontend-only, additive, on dac9fec. Desktop home voice + other omissions named but out of scope.