TopBar menus & run surfaces [UI]

The workspace TopBar's menus and the modals they open, plus the run/output surfaces reachable from the shell.

Menus & modals

flowchart LR
    TB["TopBar"]
    TB --> Settings & View & Analyze & Profiles & Reference & Debug & Right

    Settings["**Settings**"] --> s["About · Members · Agents"]
    View["**View**"] --> v["Slides · Slide regions · Theme · Time flow"]
    Analyze["**Analyze**"] --> a["Game theory · Proposition · Force · Markov"]
    Profiles["**Profiles**"] --> p["Prose · Beat · Patterns · Story settings"]
    Reference["**Reference**"] --> r["Formula · Definitions"]
    Debug["**Debug**"] --> d["API logs · System log · Regenerate embeddings"]
    Right["**Right controls**"] --> x["Gas meter ($) · Now playing · Export/Import · Learn"]

• Two wiring conventions: open a local modal (setXOpen(true)), or window.dispatchEvent('open-xxx') that a listening component handles (the page listens when a panel lives at page level).
• Learn — the badge opens LearnModal directly; the scene Learn tab and the Coverage inspector panel open it pre-scoped via open-learn-modal with a ScopeSelection.

Run & output surfaces

flowchart TB
    subgraph Capture["**CAPTURE**"]
        c["CaptureView → priors · StreamsView →<br/>commit + fold → MergesView"]
    end
    subgraph Generate["**GENERATE** (forward)"]
        g["RunBar / GeneratePanel → useAutoPlay →<br/>reasoning-graph → scenes → plan → prose"]
    end
    subgraph Scenario["**SCENARIO**"]
        rh["ReadingsView / FuturesPanel →<br/>futures-engine → sister branches"]
    end
    subgraph Review["**REVIEW**"]
        rv["SlidesPlayer · DecisionView /<br/>GameTheoryDashboard · BranchEval → ai/review"]
    end

Shipped: the room/participant model (Member/Agent/Perspective), capture-as-perspective-PRs (Streams), the war-room Merge, the Scenario card game, futures/slides, and ai/review. Not yet built: the weekly market, local encryption/PIN, Signal async capture, Cloudflare-tunnel multi-user live access. (Review-as-a-loop-phase and Butterfly were dropped.)

Learning (Quiz) layer — a purely additive, post-hoc surface: per-scene MCQ banks (ai/learning) aggregated/scoped by lib/core/learning/quiz; the three surfaces above read the same scene.questions. A Curriculum layer organises the bank into a reorganisable Topic tree, with per-member spaced-repetition Coverage on top (the Coverage tab + TreeView).