Upstream sync to fee10def1 (232 commits) with ancestry repair

[wizzoapp]

Syncs fork main with upstream pingdotgg/t3code main @ fee10de.

⚠️ LAND WITH A MERGE COMMIT ONLY. NEVER SQUASH. This PR first records the
previously-absorbed upstream ancestry (content-neutral merge of 1a003e3,
tree byte-identical, verified) and then merges the true upstream increment
(232 commits, 818 files, 42 real conflicts resolved). Squashing destroys the
ancestry repair and re-creates the ~488-file phantom conflict problem for
every future sync.

Resolution doctrine applied (per the Aug-11 sync learnings): upstream wins on
presentation/architecture (incl. class-wide restore of upstream's preview MCP
toolkit), fork wins only on data-scoping/isolation; whole assertion blocks
swept; migrations renumbered onto the fork tail; lockfile regenerated from
resolved manifests; scanner display-copy markers verified surviving with the
N-1 matrix; CI stays on fork runner labels with explicit budgets (new
upstream mobile-fingerprint job repointed to ubuntu-24.04, 60m budget).

Merge commits used the audited FACTORY_SKIP path per factory POLICY's
upstream-sync exception (supervisor ruling on record); all other commits went
through the normal gate. Focused tests, typechecks, scanner checks, and
structural audits green locally; CI is the full gate.

🤖 Generated with Claude Code

[wizzoapp]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d723c85eda

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Grant provider sessions the preview capability

For every real provider MCP credential, this guard rejects all preview_* calls because McpSessionRegistry.issue still creates the capability set as only new Set(["thread-management", "notification"]) (apps/server/src/mcp/McpSessionRegistry.ts:348). The tests manually construct scopes containing "preview", but Codex, Claude, Cursor, Grok, and OpenCode sessions never receive it, so the newly registered collaborative-browser tools always return PreviewAutomationUnavailableError before reaching the broker. Add preview when issuing provider-session credentials (and cover the issued scope in the registry test).

AGENTS.md reference: AGENTS.md:L71-L71

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Partition pull-request diff caches by audience

When private and factory-scoped clients share a server, a private request can populate this server-wide diff cache under a key containing only the epoch and PR coordinates. Factory credentials retain AuthOrchestrationReadScope, and /api/pull-requests/diff checks only that scope (apps/server/src/pullRequest/http.ts:18), so an identical factory request can receive the cached private patch without rerunning requireProject and its projection audience filter. Include the caller's audience ceiling in both the normal and stale-cache keys, or explicitly reject factory callers on the HTTP route.

Useful? React with 👍 / 👎.

[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	11.6 KiB	13.0 KiB	+1.3 KiB (+11.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.7 KiB	6.6 KiB	+850 B (+14.5%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	6.4 KiB	+511 B (+8.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	51.6 KiB	53.7 KiB	+2.1 KiB (+4.1%)	66.4 KiB	✅
Codex	Live turn messages	17	17	0 (0.0%)	21	✅
Claude	Total thread wire	11.6 KiB	12.9 KiB	+1.3 KiB (+11.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.8 KiB	6.5 KiB	+814 B (+13.8%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	6.4 KiB	+492 B (+8.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	52.4 KiB	54.5 KiB	+2.1 KiB (+4.0%)	66.4 KiB	✅
Claude	Live turn messages	17	17	0 (0.0%)	21	✅

Baseline: 3ef7cf1 · PR result: e4346ba · Source CI: failure

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: 102.2 KiB
• Claude decoded thread snapshot: 102.9 KiB

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

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e4346bad48

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Stop the loading ghost animation from looping

When pull-request list, detail, or activity data remains pending—particularly for a slow or offline remote environment—this infinite animation continues scheduling opacity-compositing updates indefinitely. Even with stepped keyframes, that leaves recurring GPU work active for every mounted loading ghost, contrary to the repository's explicit prohibition on continuously repainting animations; make this placeholder static or finite instead.

AGENTS.md reference: AGENTS.md:L136-L142

Useful? React with 👍 / 👎.