fix(cli): prevent desktop session transcript bleed from unscoped events

[konsisumer]

What does this PR do?

Stops the desktop renderer from attributing unscoped turn events to whichever chat is currently focused. Desktop now treats transcript-mutating and prompt-blocking gateway events as malformed unless they include session_id, which prevents live turns from another session/window from bleeding into the visible transcript. The backend already stamps these per-session events explicitly, so tightening the client-side routing matches the current transport contract.

Related Issue

Fixes #49106

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• apps/desktop/src/lib/gateway-events.ts: require session_id for session-scoped desktop gateway events (message.*, reasoning.*, tool.*, prompt requests, status/review events) while keeping truly global events unscoped.
• apps/desktop/src/lib/gateway-events.test.ts: update regression coverage to assert that unscoped session-scoped events are dropped instead of being attached to the focused chat.

How to Test

1. Open two Hermes Desktop chat windows/tabs on different active sessions.
2. Run turns in both sessions and switch between them while they are still live.
3. Verify a session only updates from events that carry its own session_id; malformed unscoped turn events are ignored instead of appearing in the focused transcript.
4. Local validation run for this change:
   • "$VIRTUAL_ENV/bin/python" scripts/check-windows-footguns.py --diff HEAD~1
   • git diff --check HEAD~1..HEAD
   • /opt/homebrew/bin/timeout -k 30 480 sh -c '"$VIRTUAL_ENV/bin/pytest" tests/ -q -x --timeout=60 "$@"' sh currently stops early on unrelated pre-existing failure tests/acp/test_approval_isolation.py::TestAcpExecAskGate::test_interactive_env_var_routes_to_callback
   • Workspace JS tests could not be executed in this checkout because node_modules are not installed

What platforms tested on

• macOS 26.5.1 on darwin-arm64 (local)

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 26.5.1 (darwin-arm64)

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

[alt-glitch]

Related: #47743 (resolve-session-id refactor across desktop hooks), #48281 (pin unscoped stream events to their originating session), #42359 (merged: don't drop the focused chat's own unscoped stream), #49106 (the open issue this fixes).

These are three open competing approaches to the same cross-session transcript-bleed symptom in apps/desktop/src/lib/gateway-events.ts, with different mechanisms: this PR widens gatewayEventRequiresSessionId to require session_id for all session-scoped events and DROP unscoped ones (reversing the #42178 behavior after re-verification), whereas #47743/#48281 RESOLVE/PIN the correct session for unscoped events instead of discarding them. Not a duplicate — maintainers should pick the preferred mechanism.

[alaamohanad169-ship-it]

Acknowledged: noted competing PR(s) #47743, #48281, #42359, #49106, #42178, #47743, #48281. My approach differs and is complementary, not mutually exclusive. Open to consolidating if reviewers prefer one direction.

[konsisumer]

Thanks for the analysis. I verified locally that current main already includes 28f1590b7 (fix(desktop): stop background session messages bleeding into the active transcript), so the desktop symptom this PR targets is already addressed there. This branch instead changes apps/desktop/src/lib/gateway-events.ts to drop unscoped foreground message.*/tool.*/prompt events again, which conflicts with the post-#42359 behavior that kept those events from swallowing the live answer. Given the remaining discussion is about the web /api/events sidebar path rather than the desktop path this PR edits, this PR should be closed.