chore(upstream): sync marcode through 1a003e383ac6

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[maarco]

What Changed

Daily upstream sync. Merges pingdotgg/t3code@1a003e383ac6b10258b8100c2617d938c4f06c69 into main with git merge --no-ff.

• Merge commit: 123fe314254c86e894ccef7a2414c8aca04d0855
• Parents: df91c6c4 (Marcode main) + 1a003e38 (upstream)
• Merge base: ba9c9ae81dce — the SHA the previous sync (chore(upstream): sync marcode through ba9c9ae81dce, port workspace tree onto the new sidebar #18) landed

Two upstream commits:

SHA	Subject
963ebf5b	ci: label-gated hosted-web preview deploys (pingdotgg#5465)
1a003e38	Add cross-platform mobile usage dashboard (pingdotgg#5743)

18 files: 890 insertions, 6 deletions. apps/web/src/usage/{usageFormat,usageMerge,usageMerge.test}.ts move to packages/shared/src/ so web and mobile share them; a new apps/mobile/src/features/usage/ screen consumes them; .github/workflows/web-preview.yml is new.

Why

Conflicts: none. The merge applied cleanly with zero conflicted paths — no --ours/--theirs, no reset, restore, stash, clean, or force push was used at any point. Because a clean merge is not automatically a safe merge, the sync was audited for fork-boundary breaks that produce no conflict marker.

Fork-boundary audit

Check	Result
apps/web/src/usage/** → packages/shared/src/** rename left dangling importers	No. Every importer resolves to @t3tools/shared/usageMerge / @t3tools/shared/usageFormat; no stale apps/web/src/usage reference anywhere in the tree, including docs and skills.
Marcode customizations lost in the moved files	No. Byte-diffing Marcode's df91c6c4:apps/web/src/usage/* against HEAD:packages/shared/src/* shows only upstream's own additions (one @effect-diagnostics pragma, one .ts import specifier). Marcode had no local edits in those files.
Marcode mobile branding preserved in merged hotspot files	Yes. SettingsRouteScreen.tsx keeps "Marcode Connect", "Marcode Account", and "Marcode works locally without signing in"; Stack.tsx keeps "Set up Marcode Connect". Upstream's new SettingsUsage route and settings row merged in alongside them.
New upstream mobile files carry unbranded "T3" user-visible copy	No. apps/mobile/src/features/usage/** and apps/mobile/src/state/usage.ts contain no product-name strings.
Upstream identity Marcode renames (T3CODE_HOME, t3code.service, ~/.t3, t3code:* storage keys, boot-service naming) introduced by the merge	None. The merged content reads no env vars, no storage keys, and no URLs. Tree-wide re-check confirms MARCODE_HOME (no T3CODE_HOME exists), marcode.service as the unit file, and WORKTREE_BRANCH_PREFIX = "marcode".
Compatibility identifiers kept upstream-shaped	Yes, deliberately: T3_BOOT_SERVICE_UNIT, T3_SERVICE_LAUNCHER_CONTEXT, @t3tools/* package names, and the t3 CLI name (which web-preview.yml references as t3 pair --tailscale).
Design tokens the new mobile screen depends on	font-t3-medium / font-t3-bold exist in apps/mobile/global.css; IconChartBar and the chart.bar.xaxis SF-symbol mapping were added by the same upstream commit.
Upstream deleted or rewrote a test pinning Marcode behavior	No. The only test change is usageMerge.test.ts moving with its module; no replacement pin was needed.
Coupled vendored subtrees (pnpm-workspace.yaml → .repos/effect-smol, infra/relay/package.json → .repos/alchemy-effect)	Neither source moved, so no sync:repos companion change is required.

Hotspot decisions

• apps/mobile/** (owner: mobile) — combined, not chosen. Upstream's usage dashboard is taken in full; Marcode's branding strings in SettingsRouteScreen.tsx and Stack.tsx are retained on top.
• packages/shared/package.json — combined. Upstream's two new subpath exports sit alongside Marcode's ./fractional-rank export. Stays barrel-free per AGENTS.md.
• No other hotspot path in .github/upstream-sync.yml was touched: nothing in the unified workspace tree, floating editor, sidebar, right panel, branding, provider runtime, or persistence migrations.

One thing to know about .github/workflows/web-preview.yml

Taken from upstream unchanged, on the "take their refactor" principle — rewriting it would only manufacture conflicts in future syncs. It is inert for Marcode: it runs only on same-repo PRs explicitly labelled preview:web, it is not one of the four required checks, and without VERCEL_TOKEN / VERCEL_ORG_ID / VERCEL_PROJECT_ID it exits with an explicit "missing secrets" error rather than deploying anything. Its Deploy web preview job already reported skipped on this PR, which is the expected behaviour. If Marcode never wants hosted-web previews, deleting the file is a one-line follow-up; leaving it costs nothing until someone applies the label.

UI Changes

A new mobile Usage screen (Settings → Usage) with a daily provider-stacked chart — native Swift Charts on iOS, a View-based fallback elsewhere. Not captured live: this sync ran headless in a Linux container with no iOS Simulator or Android Emulator, so the screen is verified by typecheck and static analysis only. Per docs/operations/upstream-sync.md, drive the real mobile surface before taking this out of draft.

Verification

Local checks were scoped to what the merge touched, per AGENTS.md — the full workspace suite was not run locally.

Command	Result
npx vp test run packages/shared/src/usageMerge.test.ts	9/9 pass
npx vp test run apps/web/src/components/usage/UsageProviderChart.test.ts	9/9 pass
npx vp test run apps/mobile/src/lib/adaptive-navigation.test.ts	7/7 pass
npx vp run --filter @t3tools/shared typecheck	exit 0
npx vp run --filter @t3tools/web typecheck	exit 0
npx vp run --filter @t3tools/mobile typecheck	exit 0
npx vp fmt --check on all 18 changed files	all correctly formatted
npx vp lint on apps/mobile/src/features/usage, apps/mobile/src/state, apps/mobile/src/features/settings, apps/web/src/components/usage, apps/web/src/state, packages/shared/src	exit 0
node scripts/mobile-native-static-check.ts	exit 0

Post-merge audit: zero unmerged paths, zero conflict markers, no .bak artifacts, clean worktree.

CI

All four required checks are green on 123fe314, in run 31320792146, attempt 2:

Job	Result
Check	✅ success
Test	✅ success
Mobile Native Static Analysis	✅ success
Release Smoke	✅ success

Attempt 1 shows as cancelled and should be ignored. Its Test job ran the suite in 7m25s — matching the 8m baseline on main — but the check-runs API kept reporting in_progress long after the job had settled, so it was cancelled and re-run on the assumption the runner had hung. Nothing about the diff was involved, and attempt 2 is the clean record.

Not verified: live mobile runtime (no simulator/emulator in the sync container). swiftlint / ktlint / detekt were skipped locally as not installed, but CI's Mobile Native Static Analysis job ran them and passed.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes — mobile UI could not be driven in this headless container; see UI Changes
• I included a video for animation/interaction changes — same reason

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.3 KiB	−28 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−29 B (−0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+12 B (+0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+11 B (+0.2%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.8 KiB	+1 B (+0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: df91c6c · PR result: 123fe31 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.