SimplrSSH - Accessibility AAA - 2026-10-16

[bluenevus]

Findings (WCAG 2.2 AAA 1.4.6 enhanced contrast; 1.4.11 focus indicator)

Source: main/include/ui_view.hpp theme tokens (all colors are centralized there). Measured contrast ratios:

• DIM text (hints, details, placeholder, chat info lines): 6.15:1 on BG, 5.62 on SURFACE, 4.95 on DIVIDER (<7). Affects every screen on both boards.
• DANGER text (battery low): 5.65 / 5.16 / 4.54 (<7).
• OK/WARN on SURFACE: 6.81 / 6.85 (<7).
• White text on ACCENT fill (selected row, user chat bubbles, pressed side-panel button): 3.75:1; selected-row dim text (SELECTED_DIM): 2.76:1; mark text drawn in ACCENT on white: ~3.7. Severity High: the selected row is the focus indicator users rely on.

Plan / changes

• New token theme::ACCENT_BG (0x0D47A1) for fills behind white text; ACCENT (bright, 5.05:1 vs BG) stays for borders/focus outline only. Used in ui_view.cpp (selected row, marks), chat_app.cpp (user bubble), ssh_terminal.cpp (pressed button).
• Brightened DIM 0xB1BAC4, DANGER 0xFFA198, OK 0x56D364, WARN 0xE3B341; SELECTED_DIM 0xE6EDF3.
• New guard Contrast in test/guards/test_firmware_rules.py enforces >=7:1 for all text/fill pairs and 3:1 for the focus outline; rule noted in AI_RULES.md.

Gap analysis (final)

• Theme contrast fixes + guard: done.
• Selected row remains visible as focus indicator (dark-blue fill + bright accent border + white text): done.
• Not done / not verifiable here: on-device visual check on T-Deck and Tab5 (no hardware); firmware build and RAM delta (no ESP-IDF in this environment; change is constants only, no RAM impact expected); SAST (not run locally).

Deferred (need product decision or hardware)

• Terminal palette (term_view.cpp: green on black, ANSI colors such as 0x666666/0xCD3131/0x2472C8 on black fall below 7:1) - those are server-chosen colors; a 7:1 clamp changes terminal fidelity.
• Target size 44x44 (2.5.5) on the T-Deck's 320x240 touchscreen rows/buttons - needs layout review on hardware.
• Auto-dismissed notices / timeouts audit (2.2.3) and font-size legibility review - not covered in this run.

Verification

• cd test/guards && python3 -m unittest test_supply_chain test_firmware_rules test_rules_file test_ci_scripts - passes (incl. new Contrast test).
• SKIP_SAST=1 scripts/check.sh - guards pass; host cmake step fails in this sandbox (mbedcrypto dev library missing), so host unit tests and SAST were not run locally; CI will run them.
• ruff check test: 11 pre-existing findings, same count before and after this change.

🤖 Generated with Claude Code

https://claude.ai/code/session_01MhbixBmwdwHL69hVrAaDPb

---

Generated by Claude Code