feat(web): mark sessions linked to a Mattermost chat

[tusharbhardwaj-bk]

Problem

The Linear/Mattermost bridge binds a session to a Mattermost thread so a human can steer it from chat, but that binding lived only in the bridge's own Postgres (mattermost_threads). Inside T3 nothing recorded it — the sidebar row looked identical whether or not replies in chat were landing in the session.

Fix

Adds mattermostThreadUrl to thread metadata, mirroring the existing linearIssueUrl manual tag end to end:

• packages/contracts — the field on OrchestrationThread, the thread shell, the thread.meta.update command and its thread.meta-updated payload, plus a threadMattermostLink capability so an older server hides the affordances instead of failing on click.
• apps/server — decider passthrough, both projectors, the projection_threads column via migration 1023, and every snapshot query.
• packages/client-runtime — reducer passthrough.
• apps/web — the badge itself.

The Mattermost mark renders in the experimental sidebar's bottom-right lane, immediately left of the provider icon, alongside the priority badge and owner avatar — the lane the request asked for.

Notes

• The badge is a labelled span, not a link. It sits inside the row's <button>, where a nested <a> is invalid HTML and swallows row selection. Opening the conversation is a context-menu item instead.
• Reverse states are covered: link, change, open, and remove all appear on the row menu, so a stale binding can always be cleared by hand.
• The mark is the official Mattermost brand icon (docs/site/static/img/brand/icon-denim.svg from mattermost/mattermost), recoloured to currentColor — brand denim disappears in dark mode.
• The bridge needs no new T3 endpoint: it already speaks thread.meta.update, so it can set this field on attach/bind. That bridge-side change is a follow-up in t3-linear-bridge-next and is not in this PR — until it lands, the badge lights up for links tagged by hand.
• Fork surface: all new UI lives in fork-owned files (apps/web/src/components/sidebar/); every hunk in an upstream-owned file is a marked T3-CUSTOM(expbkt3) addition beside its linearIssueUrl twin. scripts/check-fork-markers.ts passes.

Verification

Typecheck clean in packages/contracts, packages/client-runtime, apps/web, apps/server (0 errors).

Targeted tests, all passing:

• apps/web/.../MattermostThreadBadge.test.tsx — 6 new (resolver: channel/DM/permalink labels, blank + unparseable + javascript: rejection; badge: label, sizing, svg)
• apps/server/.../decider.priority.test.ts — 2 new (set/clear, and omission leaves a live binding untouched)
• apps/server/.../1023_ProjectionThreadsMattermostLink.test.ts — 1 new (column added)
• Existing suites updated for the new field and re-run green: projector.test.ts (11), ProjectionSnapshotQuery.test.ts (21), threadReducer.test.ts (33)

No screenshot yet — the badge needs a thread carrying the field, which arrives with the deploy to expbkt3. I'll attach one from expbkt3.dev once it's serving this build.

Model: Claude Opus 5 (1M context), harness: T3 Code.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ No successful main baseline artifact is available yet. This run establishes the initial measurement.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Codex	Thread snapshot wire	—	7.2 KiB	—	7.3 KiB	✅
Codex	Live turn WebSocket wire	—	6.4 KiB	—	7.8 KiB	✅
Codex	Live turn WebSocket decoded	—	55.7 KiB	—	66.4 KiB	✅
Codex	Live turn messages	—	11	—	21	✅
Claude	Total thread wire	—	13.6 KiB	—	15.1 KiB	✅
Claude	Thread snapshot wire	—	7.2 KiB	—	7.3 KiB	✅
Claude	Live turn WebSocket wire	—	6.4 KiB	—	7.8 KiB	✅
Claude	Live turn WebSocket decoded	—	56.5 KiB	—	66.4 KiB	✅
Claude	Live turn messages	—	11	—	21	✅

Baseline: unavailable · PR result: 317f7ce · 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: 110.4 KiB
• Claude decoded thread snapshot: 111.1 KiB

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