Docs: fix ornament-pattern legibility in arabic/indian/andalus/terminal themes

[testtest126]

Summary

The owner reported that text is hard to read against the background motif patterns in the indian, arabic, El-Andalus, and terminal themes on the-fleet pages. Confirmed and measured — not eyeballed.

Root cause: the four flagged themes all use small-repeat, high-density ornament tiles (hex6 66px, jaali 64px, star8 72px, scan 5px), versus the two themes that were never flagged (ensoFaint 150px, vine 130px). Pattern spatial frequency — not raw opacity — is what competes with glyph edges. Terminal's scan tile in particular repeated every 5px, close enough to a text stroke's own width to alias against it regardless of its already-low .1 opacity.

Fix, applied identically to docs/the-fleet/index.html, docs/the-fleet/night-two/index.html, and docs/architecture/index.html (all three share the same theme engine):

• Cut --orn-opacity for the four themes: arabic .16→.07, indian .17→.07, andalus .17→.07, terminal .1→.05.
• Widened the terminal scan tile 5×5 → 6×10, doubling the vertical gap between scanlines.
• Added a ground-colored text-shadow halo, scoped to just these four themes via :root:is([data-theme="arabic"], ...), on the prose that has no opaque card behind it — section ledes, ledger rows, standdown/footer text on the-fleet pages; section intros/headings/provenance line on architecture. Everything inside .role/.chan/.callout/.manifest/.laws-sec (or .col/.stage/.surface on architecture) already sits on a solid --ground-2/--surface and didn't need it.

Measured, not eyeballed: worst-case contrast (dim body text against the pattern's own color blended at full local coverage) went from 5.6–7.2:1 before to 6.1–8.5:1 after, and to 6.5–9.4:1 at the glyph edge under the halo. All four themes now clear WCAG AA (4.5:1) with real margin — they were technically already passing on raw pixel-blend math, which is why the fix targets pattern density, the actual mechanism, not just a contrast-ratio number.

Verification

Visual before/after confirmed for all four flagged themes via a static isolation harness (same CSS/SVG/markup, headless Chrome screenshots) — the pattern is now visibly quieter directly behind prose while keeping its per-theme motif intact everywhere else (margins, cards). Also spot-checked the three untouched themes (default/japanese/codex): their computed CSS is provably unaffected, since both the opacity overrides and the halo rule are scoped to the four flagged themes' own data-theme selectors (confirmed via diff — no lines changed inside their theme blocks).

Not run / why

This is a docs-only CSS/SVG change — git diff --stat confirms zero Swift/server files touched. I did not run the ChessKit/server/iOS test suites locally since nothing in this diff can affect Swift build/test outcomes; the required ChessKit tests / Server tests CI checks will still run and gate the merge per branch protection as usual.

Not security-sensitive (no auth/token/crypto/session code). Does not touch ChessKit/Sources/ChessOnline/Messages.swift or any other shared file. Does not touch #116 or #185.

🤖 Generated with Claude Code

[testtest126]

Merge request: PR #191 @ a3f29eb
Verified locally: docs-only change, zero Swift/server files touched (confirmed via git diff --stat) — ChessKit/server/iOS suites not run locally since nothing in this diff can affect them; required CI checks will still gate the merge as usual.
Security-sensitive: no
App-touching: no
Shared files: none touched, not Messages.swift

Fix for the owner's report that indian/arabic/El-Andalus/terminal themes had illegible text against the background ornament patterns — reduces ornament opacity + tile density for those four themes and adds a scoped text-shadow halo behind unprotected prose, applied identically across docs/the-fleet/, docs/the-fleet/night-two/, and docs/architecture/. Full before/after visual verification and WCAG contrast math in the PR description. Does not touch #116 or #185.

[testtest126]

Follow-up pushed @ e633434 — this voids any prior CI/review state seen at the old head (a3f29eb).

What changed: the Assayer's independent review caught that --bone-faint (used in .sig/footer/.inc .refs) was clearing WCAG AA only via the text-shadow halo, not on color math alone — flagged terminal at ~2.98:1. Re-measuring myself found all four fixed themes were actually short on raw color math (arabic 4.24:1, indian 4.25:1, andalus 3.66:1, terminal 2.98:1), not just terminal.

Fix: lifted --bone-faint per theme (blended toward each theme's own --bone-dim, already proven safe, rather than picking arbitrary new colors) until worst-case contrast clears 4.5:1 with margin. Re-verified by parsing the actual edited source values, not hand-copied numbers:

theme	old	new	worst-case now
arabic	#7d7f9c	#8688a3	4.77:1
indian	#96806a	#a0886f	4.75:1
andalus	#8d8267	#a3967a	4.77:1
terminal	#4f6654	#6e8772	4.76:1

All four clear AA on the color math alone now — halo is a bonus, not load-bearing. Applied to docs/the-fleet/index.html and docs/the-fleet/night-two/index.html only; docs/architecture/index.html needs no change since its equivalent token (--ink-soft) is numerically identical to the-fleet's --bone-dim (already safe at 6.08–8.49:1), not --bone-faint — no third faint tier is exposed to the pattern there.

Still docs-only, still not security-sensitive, still doesn't touch #116 or #185. No orchestrator-approval posted.