Prepare trust-focused TestFlight beta

[cbusillo]

Objective

Prepare Context Panel for a trust-focused TestFlight beta. The beta should prove the native app, widget, login item, provider adapters, local storage, and permission model behave predictably on real machines before adding more cleverness.

The guiding promise is: a tester can set up Code/OpenAI, Claude, and Gemini; the widget keeps updating; failures are understandable; stale data is preserved only when that is safer than deleting useful state; and every ready-to-test claim is backed by the canonical /Applications/Context Panel.app runtime receipt.

Finish Line

Context Panel is ready for TestFlight when the must-fix child work is complete, packaging validation passes, and a fresh canonical runtime receipt shows baseline=OK for the installed app, widget extension, URL handler, and refresh login item.

Current Status

State: Active. Packaging validation #71 is complete; reset priming remains deferred in #75.
Next action: Break this umbrella into the scoped child issues below and start with setup clarity plus provider/account detail polish.
Blocked by: No current packaging blocker. Reset priming remains related/deferred in #75.
Last verified: 2026-05-16 during cross-repo issue audit; stale blocker wording for completed #71 was clarified.

Scope

• In: TestFlight readiness, setup clarity, provider/account detail, refresh/stale/failure behavior, diagnostics/status language, runtime validation, and App Store/TestFlight packaging readiness.
• Out: reset priming implementation, notifications, deep history/charts, advanced widget customization, smart recommendations, and new provider reverse-engineering unless current supported flows break.

Acceptance Criteria

• Setup flow clearly tells testers which file/source to authorize for each provider and what success/failure means.
• App detail views expose combined and per-account OpenAI limits, including email, plan, 5-hour, weekly, and additional limits where useful.
• Background refresh settings persist, login item state follows saved settings, and manual/background refresh produce compatible snapshot shapes.
• Provider failures are plain-language and actionable; raw HTTP/status details are kept to diagnostics.
• Stale and failed data handling is deliberate: preserve useful last-known data when the provider has no fresh data, but do not overwrite fresh sibling data with stale limits.
• Runtime validation remains codified: before any ready-to-test/TestFlight handoff, scripts/context-panel-runtime-baseline.sh check or install --launch completes with baseline=OK.
• TestFlight packaging/export/upload path is validated separately.

Relationships

• Related: Validate TestFlight packaging and App Store icon separately #71 for TestFlight packaging and App Store icon validation.
• Related/deferred: Finish safe reset priming for supported providers #75 for safe reset priming.
• Related: Stabilize Context Panel runtime validation #65 for runtime validation stabilization.
• Related: Restore Context Panel settings and account management surface #67 for settings and account management surface.
• Related: Fix Gemini access from signed sandbox runtime #68 for Gemini signed sandbox access.
• Related: Deduplicate stale provider account IDs in snapshots #69 for stale provider account ID snapshot cleanup.

Child issues to create/link:

• Setup clarity and permission repair for provider files.
• OpenAI provider/account detail polish.
• Refresh, stale-data, and login-item trust pass.
• Diagnostics and user-facing state language pass.

Validation

Use these as the release-readiness baseline:

swift test --filter 'ProviderConnectorTests|SnapshotStoreTests|FastModeForecastTests|WidgetSnapshotTests|AccountConfigurationStoreTests|ResetPrimerPlannerTests'
scripts/context-panel-runtime-baseline.sh install --launch

For final TestFlight packaging, use the release/package scripts and validate the exported/uploaded artifact per #71.

Decisions

• Ship a trust beta before adding more intelligence.
• Keep reset priming deferred until it can be cheap, account-safe, and unsurprising.
• The widget should remain combined/glanceable; per-account detail belongs in the app.
• The canonical runtime remains /Applications/Context Panel.app; DerivedData/worktree bundles are not valid app/widget/provider test runtimes.

Open Questions

• Do we want one TestFlight build with only Chris as tester before inviting outside users?
• How much diagnostics detail should be visible by default versus hidden behind a Troubleshooting section?
• Should the default refresh interval remain 5 minutes for beta, or move to 10 minutes for less background activity?

[cbusillo]

Created child issues for the TestFlight trust-beta plan:

• Clarify provider setup and permission repair for TestFlight #77 Clarify provider setup and permission repair for TestFlight
• Polish OpenAI provider and per-account limit detail for TestFlight #78 Polish OpenAI provider and per-account limit detail for TestFlight
• Harden refresh, stale-data, and login-item trust before TestFlight #79 Harden refresh, stale-data, and login-item trust before TestFlight
• Tighten status language and diagnostics for TestFlight #80 Tighten status language and diagnostics for TestFlight

Existing related/blocking work:

• Validate TestFlight packaging and App Store icon separately #71 remains the packaging/TestFlight export validation track.
• Finish safe reset priming for supported providers #75 is explicitly deferred/later for reset priming. It should not block the first trust beta unless we decide priming is required for the product promise.
• Stabilize Context Panel runtime validation #65, Restore Context Panel settings and account management surface #67, Fix Gemini access from signed sandbox runtime #68, and Deduplicate stale provider account IDs in snapshots #69 are related historical/runtime/provider tracks that should be swept before closing this umbrella so stale plan state does not mislead a future session.

[cbusillo]

2026-05-14 readiness snapshot:

Passed locally on branch fix/widget-refresh-regression:

• swift build
• swift test: 146 tests passed
• scripts/commit-gate.sh
• scripts/context-panel-runtime-baseline.sh check: baseline=OK, app/widget/URL handler/refresh agent all resolve to /Applications/Context Panel.app, signed for team MM5YXC7T6E, current snapshot savedAt=2026-05-14T15:43:50Z

Issue updates:

• Closed Fix Gemini access from signed sandbox runtime #68: Gemini signed-runtime access is validated.
• Closed Stabilize Context Panel runtime validation #65: runtime validation is now codified and receipt-backed.
• Updated Finish safe reset priming for supported providers #75: reset priming remains deferred until cheapest/account-selectable execution is proven.
• Clarify provider setup and permission repair for TestFlight #77/Polish OpenAI provider and per-account limit detail for TestFlight #78/Harden refresh, stale-data, and login-item trust before TestFlight #79/Tighten status language and diagnostics for TestFlight #80 remain active polish/hardening tracks before TestFlight.

Known non-code-gate gap: JetBrains inspection tool was unavailable in this session, so no IDE inspection result is recorded.

[cbusillo]

Readiness snapshot after the refresh/login-item pass:

Closed as completed:

• Stabilize Context Panel runtime validation #65 runtime validation stabilization.
• Deduplicate stale provider account IDs in snapshots #69 duplicate stale provider account IDs in snapshots.
• Harden refresh, stale-data, and login-item trust before TestFlight #79 refresh/stale-data/login-item trust.

Current evidence:

• scripts/commit-gate.sh passed; full Swift suite reported 148 tests passing.
• Installed refresh-agent --refresh-once exited 0 and wrote savedAt 2026-05-14T16:22:10Z.
• scripts/context-panel-runtime-baseline.sh check passed with baseline=OK for /Applications/Context Panel.app, PluginKit widget registration, contextpanel:// handler, and the running refresh login item.
• Live snapshot from the installed runtime has both OpenAI accounts healthy, Gemini healthy, and Claude stale.

Still open before TestFlight:

• Clarify provider setup and permission repair for TestFlight #77 setup/permission repair clarity.
• Polish OpenAI provider and per-account limit detail for TestFlight #78 OpenAI provider/per-account UI polish.
• Tighten status language and diagnostics for TestFlight #80 status/diagnostics language polish.
• Validate TestFlight packaging and App Store icon separately #71 packaging/export/upload validation.
• Finish safe reset priming for supported providers #75 reset priming remains intentionally deferred.

[cbusillo]

Additional progress after #77 setup-state patch:

• Settings now recognizes shared-Keychain imported auth JSON as a saved provider setup state for Code/Codex/Gemini.
• Claude row wording was adjusted so it does not imply a normal auth-file picker.
• Canonical install gate passed again: baseline=OK for /Applications/Context Panel.app with fingerprint 4ba8627d55aa690ef2497d4c9fc519cbfa541ee15e1133829712350076481503.
• Snapshot savedAt 2026-05-14T16:31:43Z; OpenAI two accounts healthy, Gemini healthy, Claude stale.

Still open before TestFlight: #77 visual/copy completion, #78 OpenAI detail polish, #80 diagnostics/status language, #71 packaging/export/upload.

[cbusillo]

Closed #78 after checking the current OpenAI provider route and live installed snapshot. The app now exposes per-account OpenAI email/plan plus 5-hour and weekly main limits in the app, while Spark/additional limits remain separate and the widget can stay combined.

[cbusillo]

Correction after auto-review on the setup/keychain path:

• Fixed keychain read failure fallback so the app can still use the bookmark/direct file path when the credential cache is unavailable.
• Fixed legacy document-scoped bookmark resolution to keep using the bookmark store URL as the relative base.
• Revalidated with focused tests: 38 tests passed.
• Reinstalled canonical runtime: scripts/context-panel-runtime-baseline.sh install --launch passed with baseline=OK, fingerprint 53aa1dd72278f1d6b1e290595392c61f4c73489d0f6f93ab5cc935318d9eada6.
• Current installed snapshot savedAt 2026-05-14T16:47:22Z; both OpenAI accounts healthy, Gemini healthy, Claude stale.

[cbusillo]

2026-05-14 cleanup status:

Current state after the TestFlight trust pass:

• Canonical runtime workflow is codified and passing: latest install gate showed baseline=OK for /Applications/Context Panel.app, widget registration, URL handler, and refresh login item.
• Provider setup now exists in Settings: Code/OpenAI file auth, Gemini file auth, Claude OAuth connect, background refresh settings, widget main-limit preferences, Diagnostics, and disabled/coming-soon Reset Primer.
• Closed superseded/completed planning issues: Add Claude status-line setup onboarding #18, Reset app information architecture after widget UI prototype #23, Move account setup into macOS Settings #25, Restore Context Panel settings and account management surface #67.
• Removed local scratch artifacts: .tmp/testflight-readiness-body.md, .claude/settings.local.json, and the one-off OAuth inspection script.

Still open for TestFlight:

• Validate TestFlight packaging and App Store icon separately #71 packaging/export/upload validation and App Store icon/package evidence.
• Clarify provider setup and permission repair for TestFlight #77 final provider setup/repair bad-path QA.
• Tighten status language and diagnostics for TestFlight #80 can close after the current branch is merged and the final copy is visually accepted.
• Finish safe reset priming for supported providers #75 remains intentionally deferred for safe reset priming.