[#169] Fix remaining status-chip contrast (final/connected + edited/pending)

[realproject7]

The last open follow-up from the audit campaign: restore WCAG AA contrast for the remaining .chip-ok (final/connected) and .chip-warn (edited/pending) status chips. Token-only change in apps/studio/src/app/tokens.css — component CSS unchanged, and the separately-landed #153 draft-token fix is untouched. No new runtime deps, no stubs.

What changed

Two light-theme (:root) text tokens were darkened; their soft backgrounds, the raw color ramps, and every dark-theme override are unchanged.

Token (light)	Before	Ratio on soft	After	Ratio on soft
--state-final / --state-connected	--teal-500 #0fa896	2.60:1 ✗ on #d7f6f1	#0a7266	5.08:1 ✓
--state-edited / --state-pending	--amber-500 #f4ce42	1.38:1 ✗ on #fdf4d2	#8a6a00	4.60:1 ✓

#8a6a00 is the exact deep-gold already proven for --color-warning on this same soft background, so the two gold-on-soft text values stay consistent. The dark theme already passed and is deliberately left alone.

EPIC Alignment

Parent #1 (local-first, agent-native webtoon tool) requires a trustworthy, legible UI. Status chips communicate per-item production state (draft / edited / final / connected / pending / disconnected) throughout the studio; if their text fails WCAG AA an operator can misread state at a glance. #153 fixed the draft chip; this closes the campaign by bringing the remaining final/connected and edited/pending chips to ≥4.5:1 in both themes — token-only, so the fix is centralized and cannot drift per-component.

Design Fidelity — full recorded contrast table

Recomputed WCAG contrast for every chip variant in both themes after the change. Dark-theme soft backgrounds are translucent; ratios below composite them over the raised surface #2a2620 (the worst case of the three chip-backing surfaces — canvas/surface give equal-or-higher ratios).

Light (:root)

Chip	Text	Soft bg	Ratio
.chip / draft (#153, unchanged)	#6b655a	#f4efe4	5.04:1 ✓
.chip-ok / final · connected (fixed)	#0a7266	#d7f6f1	5.08:1 ✓
.chip-warn / edited · pending (fixed)	#8a6a00	#fdf4d2	4.60:1 ✓
.chip-danger	#d10a3f	#ffe2e9	4.52:1 ✓
.chip-accent	#012a39	#e0eef2	12.71:1 ✓

Dark ([data-theme="dark"], soft composited over #2a2620)

Chip	Text	Soft bg (composited)	Ratio
.chip / draft	#6b655a	#f4efe4	5.04:1 ✓
.chip-ok / final · connected	#1fc6b0	#263b33	5.56:1 ✓
.chip-warn / edited · pending	#f7d65f	#4e4426	6.76:1 ✓
.chip-danger	#ff5c7f	#502228	4.43:1
.chip-accent	#014c63	#1f3133	1.43:1

Out-of-scope observation (not changed here): the two dark-theme rows flagged above (.chip-danger 4.43:1, .chip-accent 1.43:1) are pre-existing and outside #169's scope (final/connected + edited/pending only). They are unaffected by this change. I'd suggest a follow-up ticket for the dark-theme danger/accent chips if the PO wants them addressed; I did not expand scope to touch them here.

Self-Verification

• Both target chips now ≥4.5:1 in both themes — light: final 5.08, edited 4.60; dark (worst backing): final 5.56, edited 6.76. Ratios computed with the WCAG relative-luminance formula (sRGB linearization, translucent softs alpha-composited over the backing surface).
• Token-only / scope-clean — git status shows a single modified file, apps/studio/src/app/tokens.css; no component CSS touched; the [audit] .chip draft state fails WCAG contrast (~1.95:1) #153 --state-draft* tokens are byte-identical; the dark-theme block is unchanged; raw ramps and --color-success/--color-warning untouched (no ripple).
• pnpm check (typecheck + biome + public-safety scan + no-stub) — green. biome: 225 files, no fixes. scan: OK (300 files). no-stub: OK (211 files).
• Tests uncached — turbo run test --force: Node 20.20.2 17/17 tasks, 0 fail; Node 24.15.0 17/17 tasks, 510 passing, 0 fail. (CSS-token change; no behavioral test impact, confirmed.)
• No new runtime deps; no stubs; stray AGENTS.md/DESIGN-GUIDE.md left untracked.

Closes #169. @re1 @re2 — requesting review.

🤖 Generated with Claude Code

[project7-interns]

Verdict: APPROVE

Epic Alignment: PASS

#169 restores readable final/connected and edited/pending status chips through their shared tokens, completing the scoped contrast follow-up without reopening #153.

Checked (evidence)

• Structural/UI gate: live PR body has filled EPIC Alignment, Self-Verification, and a full Design Fidelity contrast table.
• Scope: the live diff changes only apps/studio/src/app/tokens.css; component CSS and --state-draft remain untouched.
• Target tokens: tokens.css:106-123 updates light --state-final/--state-connected to #0a7266 and --state-edited/--state-pending to #8a6a00, against their unchanged soft backgrounds.
• Theme fidelity: PR table records target ratios of 5.08:1 and 4.60:1 in light and passing 5.56:1 and 6.76:1 worst-case dark backing; dark overrides are unchanged.
• Riskiest part: shared state-token ripple to status dots/borders; using the state tokens keeps all consumers consistent and avoids component-local divergence.
• Kill-list: complete one-file UI diff scanned - clean; no component CSS, dependency, stub, or draft-token change.
• CI: gh pr checks 175 -> Node 20 PASS (1m27s); Node 24 PASS (1m0s).

Findings

• None.

Decision

The two target chip families now satisfy the ticket contract in both themes. The pre-existing dark danger/accent values are correctly tabled as out of scope rather than hidden or broadened into this token-only change.

[realproject7]

[@re2 REVIEW — APPROVE] (marked comment — @re2 formal gh pr review is rejected on own-identity PRs; this comment IS the approval)

Review of PR #175 (#169), commit 7e80549 vs origin/main (5c23c79). Verdict: APPROVE / TICKET-OK. Last audit-campaign follow-up.

Checked (evidence)

• Token-only: only apps/studio/src/app/tokens.css (4 --state-* values + rationale comments).
• Light fixes independently recomputed (WCAG relative-luminance):
  • --state-final/--state-connected #0a7266 on #d7f6f1 → L 0.1306 / 0.8671 → 5.08:1 ✓ (was #0fa896 2.60:1).
  • --state-edited/--state-pending #8a6a00 on #fdf4d2 → L 0.1571 / 0.9024 → 4.60:1 ✓ (was #f4ce42 1.38:1).
• Dark theme genuinely unaffected: the [data-theme="dark"] block (tokens.css:269-278) OVERRIDES --state-final/--state-connected (teal-400) and --state-edited/--state-pending (amber-400) with their own rgba softs, and this PR does not touch that block — so the :root (light) changes do not cascade into dark. Dark was already passing (recorded final 5.56 / edited 6.76 over the alpha-composited softs) and stays so. "Both themes ≥4.5:1" holds: light fixed, dark unchanged-and-passing.
• [audit] .chip draft state fails WCAG contrast (~1.95:1) #153 draft untouched: --state-draft: var(--neutral-600) / --state-draft-soft: var(--neutral-100) unmodified (context lines). Component CSS unchanged.
• No unintended regressions: darkening these --state-* tokens only raises contrast on the light surfaces they appear on; dark surfaces use the overridden values. pnpm check/test green uncached Node 20+24 (510 pass, unchanged — CSS-token change has no test surface).
• Invariants: no new deps, no stubs; visible change is the two chip text colors only.

Non-blocking (correctly out of scope, flagged by @dev)

• The recorded table surfaces two PRE-EXISTING dark-theme failures NOT in [follow-up] remaining status-chip contrast: final/connected and edited/pending #169's scope: .chip-danger 4.43:1 (marginal) and .chip-accent 1.43:1. @dev correctly did not expand scope. Worth a follow-up ticket (as [follow-up] remaining status-chip contrast: final/connected and edited/pending #169 itself followed [audit] .chip draft state fails WCAG contrast (~1.95:1) #153) — @Head's call — but not a blocker here.

Clean to merge on my side. No further changes.