test: add accessibility and Python 3.13 evidence

[YuzeJ21]

Summary

• replace the partial packaged-browser path with a true keyboard-only installed-wheel journey using Tab, keyboard typing, Enter, and Space, with focus-specific visible-treatment and in-viewport assertions for each primary target;
• reject unchanged always-on decoration and transparent focus styling with focused-versus-unfocused computed-style regression coverage;
• add a separate CPython 3.13 compatibility job covering the complete suite, both deterministic benchmarks, wheel reinstall, dependency validation, both CLIs, and exact loopback health without renaming the existing protected jobs;
• record bounded native Chrome 200% zoom evidence and update the authoritative gap ledger;
• preserve unsupported classifications for VoiceOver, Windows desktop, Linux desktop, and accessibility conformance.

ScopeProof product and deterministic-gate source code do not change in this PR.

Exact local evidence

PR head: cee5921d064c77ba420647a1ffd4f2d72cbd3a9d

Final executable/test verification head: 1a3defcfec5f9f2231a67e8f0a4450d43a63b9b4 with tree b01e358b5aefefd7eaf8abf28e0bf7502cf962af. The following cee5921d commit only records the verified evidence in the audit document and does not change executable or test content.

• Ruff: passed.
• Complete suite: 1,956 passed, 2 intentional skips, 95.06% combined coverage.
• Focused installed-wheel browser suite: 3 passed, including regressions for unchanged decoration and changed visible focus treatment.
• Repository contracts on the PR head: 77 passed.
• Deterministic benchmark: 12 cases / 13 criteria, zero mismatches, zero must-have False Ready outcomes, zero false blockers, zero unexecuted categories.
• Comparison benchmark: 2 cases, zero mismatches.
• Installed-wheel Chromium regression: passed at 1280×720 and 390×844 with keyboard-only primary actions, focus-specific visible treatment, no external requests, no console errors, and no page errors.
• Reproducible wheel: two byte-identical 101-entry builds; SHA-256 9afb3cd5a2cce17aaa047abf1c7924fd9e74bd2faa834aab75134c6f92671069.
• Fresh Python 3.12 and genuine CPython 3.13.14 installs: dependency checks, package/version equality, both CLIs, both installed benchmarks, and exact workbench health ok passed.

Actual 200% zoom boundary

Installed Chrome 151.0.7922.77 on macOS 26.5.1 visibly reported Zoom: 200%. Device pixel ratio changed from 1 to 2 while the physical outer window stayed 3440×1328; the CSS viewport changed from 3440×1151 to 1720×575. The constructed workflow completed by keyboard only, all primary targets exposed visible focus, the fail-closed Action required result and three exports appeared, and no named control was horizontally clipped. Chrome was restored to 100%.

Responsive viewport resizing is not presented as native zoom evidence. This result is limited to the exact Chrome/macOS/display configuration above.

Unsupported evidence

• VoiceOver was installed and detectable, but two bounded attempts exposed no controllable, observable screen-reader state, speech, caption, focus-announcement, or reading-order evidence.
• No real Windows or Linux desktop environment was available. The Python 3.13 Ubuntu runner is package/runtime evidence only, not Linux desktop evidence.
• Keyboard, focus, DOM, automated browser, and native-zoom checks do not establish WCAG conformance or usability for every user.

Evidence boundary

This PR contains ScopeProof engineering evidence only. It does not prove correctness, target-repository runtime behavior, authenticated reviewer authority, customer use, production use, adoption, or accessibility conformance. It does not execute target-repository code. Static candidates remain distinct from runtime verification, criteria confirmation remains mandatory, Pydantic validation and deterministic fail-closed gates remain unchanged, and the GitHub Action remains opt-in and informational.

Stage 1 remains:

• 0/5 qualifying reviews;
• 0/3 independent practitioners;
• 0/3 public repositories;
• 0/3 independently observed under-ten-minute completions;
• 0/2 reuse-intent signals.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 03b001221f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1a3defcfec

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".