docs(plan): record the G11 follow-up as shipped, and what stays out

[mizu-jun]

What

The G11 follow-up entry in docs/plans/ui-ux-modernization-v3.md has been an open checkbox holding a list of sites since #62. It now records what shipped — #69 (terminal surfaces), #70 (modal scrim + destructive-button reds), #71 (modal dialog buttons) — split by surface, plus the two sites that stay out.

Why those two stay out (decisions, not leftovers)

• overlay/picker.rs's query/selection colours. The purple/green macro & SSH branding is intentional per its own code comments, so migrating it is a product decision about whether Nexterm keeps per-feature brand hues at all — not a token substitution.
• color_util::resolve_color's no-palette fallbacks. render_frame builds scheme_palette as an unconditional Some, so the None arm is reachable only from tests. Tokenising it would change nothing that renders. (Noted in passing: the same line makes the DesignTokens::default() branch beside it dead too — a future cleanup, not G11's.)

Three findings recorded in the plan, not just in commit messages

Each contradicts something the previous entry assumed:

1. The delete-dialog reds had drifted further than recorded — not only the focused fill ([0.498, 0.196, 0.196] vs [0.486, 0.180, 0.180]) but the resting treatment (dark red vs surface_1) and the label rule.
2. The scrim was never "one literal in four files". It was an asymmetry: five modal surfaces, only the settings panel scheme-derived, so a light scheme put a black veil behind a light panel.
3. No fixed blend strength can carry a semantic fill across all nine schemes. Nord (error hue at 0.85 → 4.42:1) and Solarized (warning used raw → 4.37:1) land at luminances where neither a near-black nor a near-white label has anything to contrast with. Hence semantic_fill stepping the blend back until the label clears the floor.

On-device backlog

Gains #69 / #70 / #71, with #70 flagged as the highest-value item in the whole list: the scrim is the one place this work changed what an existing surface looks like rather than only where its value comes from (four modals went from a black veil to surface_0, so on a light scheme from dark to light). That was reasoned about, never seen.

The section now also states what the colour work did get machine-verified — the contrast floors are pinned by tests across all nine schemes, and two of those tests fail if a hard-coded literal returns — so the line between measured readability and unverified appearance stays explicit rather than blurring into "it's tested".

Test plan

• Documentation only — no code paths touched
• Markdown nesting and continuation-line indentation match the surrounding entries
• Every PR number referenced (refactor(client): migrate the G11 hard-coded colours onto design tokens #62, refactor(client): move the terminal-surface colours onto tokens (UI/UX v3 G11 follow-up) #69, refactor(client): share the modal scrim and the destructive-button reds (UI/UX v3 G11 follow-up) #70, refactor(client): put the modal dialog buttons on tokens, readably (UI/UX v3 G11 follow-up) #71) is merged

🤖 Generated with Claude Code

[github-actions]

Coverage report

Generated by cargo llvm-cov (workspace minus nexterm-client-gpu and nexterm-i18n).