Move the fold marker left, and make the ? pane's keys the bright thing

[navbytes]

Two changes from reading a real screen of beta.2, plus a TUI-UX review of the ? pane.

1. The fold marker moved to the row's left edge

before   │ ● user: yes build it   ▸ 38 steps · ~23.5k                      ~23.5k
after    │ ● ▸38 user: yes build it                                        ~23.5k

▸38 now sits between the ● and the text: the row's own disclosure control, where an outline puts one. Two things were wrong with the trailing digest:

• The caret was nowhere near the ● it belonged to, so on a tree of folded turns nothing marked a collapsed row until you had read to the end of a preview that is usually clipped.
• ~23.5k was the row's own token column again. applyFolds rolls the hidden steps' tokens into the turn, so the two figures are equal by construction — two numbers of the same magnitude a few columns apart that always agree read as one number that has gone wrong.

foldDigest becomes foldMark (the caret and the count, on the left) and foldFlags ( 1✂ 2⚠ 1✗, at the end of the text, in the same place and glyphs a step row already uses for its own). The tokens are printed once. No unit word — a bare count beside a caret is unambiguous, and "steps" was six columns the preview wanted.

2. The ? pane

Its colour was exactly inverted

route.tsx coloured whole lines, keyed on whether the line was indented:

<text fg={l.startsWith(" ") ? t.textMuted : t.accent}>│ {l}</text>

Every line that names a key is indented — so the headings, which name no keys at all, were the one accent-coloured thing on the pane, and every key was muted. Someone opening ? to find one key had no token-shaped thing to scan for and had to read the pane at reading speed.

The pane is now a list of typed segments (key, name, label, heading, glyph, strong, text) that the route colours individually. Keys are the brightest and the only bold thing; headings accent; prose muted; the │ gutter always dim instead of taking its line's colour. helpLines() remains as the joined-string API the drift tests assert against.

Act and Views became the table they already were

Every row in them was already <key> <name> — <purpose> in a consistent order. The " — " is now two columns aligned at render time, so you can enter from either side — scan the keys if you think gm, the names if you think "merge". Widths are measured per section, so one f12 rebound in Act cannot shift Legend, and a key past 8 columns takes its own row ragged rather than pushing its whole section right.

Move deliberately keeps its packed clauses: tabulating twenty motion clauses would cost twenty rows to make vim's own keys — the ones this audience already has — the most prominent thing on the pane. It gets a dim static label column instead (rows / jumps / folds / search), as does Legend (rows / tree / cost / status / ⎇ / lanes). Those labels are static strings, so no rebind can break their alignment.

Nine lines were clipping at 100 columns with the width test green

A help row is │ inside a padding={1} box, so its budget is cols - 4 — which is what every other line on the screen uses — but the test asserted ≤ 112 against the raw string. One line clipped even at 112. The pane is now laid out for the terminal it is on: clauses drop from the right, prose clips with …, nothing wraps (§7.6), and the same 34 rows survive from 100 columns down to 44.

Smaller

• The footer names the section you are in (12–29 of 34 · Act · PgUp/PgDn scroll · q/esc back) — the pane is a three-page scroll on a short terminal and its headings scroll off. Zero extra rows.
• The legend draws its colours instead of spelling them: █ open · █ squashed · … in the branch colours, lane swatches in the lane colours. Naming a colour in muted grey cost six columns and asked the reader to take a monochrome word's word for it; showing it costs one — which is what let the pane's two longest lines fit.
• Glyphs take the tree's own colours (⚠ warning, ✗ error, ◆ ≣ ✂ accent); nothing here rewrites your transcript is bold as the safety claim it is; the *model* markdown asterisks became a bold run.

Net: 34 rows before, 34 after.

Two bugs found on the way — both caught by the pane's e2e while the unit suite stayed green

• A runtime import of @opentui/core broke the whole route. Using TextAttributes.BOLD added a runtime import the plugin had never had (peer dep only), and the host then failed to load the route at all — /tree simply never opened. The two bits are written out as constants, pinned against the real enum by a test that also asserts the built bundle imports only @opentui/solid.
• A createMemo in the temporal dead zone. Moving HELP to read size() left it above its own dependency's declaration; a createMemo body runs the moment it is created, so it threw on every render.

Tests

• test/help.test.ts — rewritten against the segment model. The drift guard is tightened in the direction that matters: every bound verb must appear inside a key segment, not merely somewhere in the line's text, since a key that leaks into a prose run is invisible again. Mutation-checked: folding one key back into a prose run fails 6 tests. Width is asserted against the row budget at four widths (96 / 76 / 56 / 40), and narrowing must never drop a row.
• test/fold.test.ts — foldMark / foldFlags, including that the marker never repeats the token column.
• test/e2e/tui.test.ts — the fold test asserts the marker is between the ● and the preview and is not a digest; the ? pane test asserts the new table shape and that the footer names a section.

bun run typecheck, bun test (462 pass) and bun run build are clean. The full pty e2e suite was still running when I opened this; I'll report on the PR if anything comes back red.

Known gap

docs/screenshots/help.png, linked from the README, now shows the old pane. It needs a fresh capture on a real terminal — I can't produce a PNG from here.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VkCWU9Ab6wYHmTke8vuVMv

---

Generated by Claude Code