Enable voice-only Aiden Live by default in 0.41.5

[sambitcreate]

Summary

• enable ordinary voice-only Aiden Live for normal Finder-launched production builds
• preserve exact 0/false incident disable values and fail closed on unknown explicit values
• keep screen capture separately gated behind AIDEN_EXPERIMENTAL_GEMINI_LIVE_SCREEN=1
• bump the release version to 0.41.5 and update the Live plan/status evidence

Verification

• npm run test:gemini-live (181 Live + 48 Computer Use tests)
• npm run type-check
• npm run lint
• npm run test:branding
• npm run build
• node --test scripts/check-macos-release.test.mjs
• npm run release:check-consumers
• local release preflight reached the expected Apple credential boundary; the credentialed GitHub release workflow remains the signing/notarization gate

[pullfrog]

Important

The default-on rollout is sound, but the emergency override does not fully fail closed for an explicitly empty value.

Reviewed changes This review covers the default Live capability gate, the separate screen-capture gate, release metadata, and the focused test updates.

• Voice-only capability gate — geminiLiveEnabled now enables Live when unset, preserves the 0/false kill switch, and rejects other non-empty values.
• Screen and action boundaries — screen sharing still requires AIDEN_EXPERIMENTAL_GEMINI_LIVE_SCREEN=1, while Computer Use retains its independent persisted enablement, readiness, and authorization checks.
• Release metadata and coverage — the package version and Live plan are updated, and the focused Live and Computer Use suites cover the new default behavior.

ℹ️ Nitpicks

• tests/e2e/fixtures.ts:755-756 still says the production candidate is acceptance-gated. The explicit AIDEN_EXPERIMENTAL_GEMINI_LIVE=1 remains harmless, but the comment should explain its current test purpose now that the production default is enabled.

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using GPT Luna ｜ 𝕏}

[pullfrog]

✅ No new issues found.

Reviewed changes Re-reviewed the changes since the prior Pullfrog review and the full PR context.

• Separated absent and explicit values — Updated geminiLiveEnabled to default on only for an undefined flag while rejecting explicit empty, whitespace-only, and unknown overrides.
• Added fail-closed regressions — Added exact assertions for empty and whitespace-only overrides while retaining the kill switch and screen-gate coverage.

  ^{｜ View workflow run ｜ Using GPT Luna ｜ 𝕏}