docs(plans): execution plans for the six open issues

[oscarlehuu]

Summary

One plan file per open issue (#102, #104, #105, #109, #110, #111) plus a
sequencing index at plans/20260809-0355-open-issues-sequencing/plan.md.

Docs only — no code, no CI config, no runtime behaviour.

What these plans do and do not contain

#102, #104, and #105 already carry complete phase plans inside their issue
bodies. These files deliberately do not copy that content — a duplicated spec
drifts from the original and the copy becomes silently wrong. For those three the
plan covers only what the issue does not: current status, cross-issue coupling,
Crew-specific risk, and ordering. The issue stays the spec.

For #109, #110, and #111 the plans are complete, including evidence gathered while
filing them.

The sequencing call

Restore the e2e safety net before building more product on it.

Desktop Smoke E2E (4) has been cancelled at the 30-minute timeout on 6 of 6
consecutive main runs since b57d26def (#95). At e41a1a6a4 it was killed after
test 106 of 250 — 144 tests never ran. Shard 1 has hard-failed since 25263120e (#96).

The three not-started feature issues all ship behaviour whose acceptance criteria
are about replay, reconnect, and restart. None of that is provable by unit tests.

Recommended order: #109 -> #110 -> finish #105/PR #108 -> #111 decision -> #104 -> #102.

Root cause included for #110

Bisected locally in an isolated worktree, single test, build verified each round:

tree	result
b57d26def (#95)	1 passed
25263120e (#96)	1 failed - 4/4 runs
25263120e with AppShell.tsx reverted	1 passed - 3/3 runs

AppShell.tsx now passes every channel id into useLiveHomeFeedActions, which
opens an app-wide live subscription per channel using the same filter
useChannelUserInput uses. The test's readiness gate
(__BUZZ_E2E_HAS_MOCK_LIVE_SUBSCRIPTION__) returns a plain boolean, so it is now
satisfied by the AppShell subscription before ChannelPane has subscribed - the
emit lands before the channel hook is listening.

The if (!conversationId) return; early-return flagged in issue #110 is not the
cause; the id falls back to event.id, which is valid, so the guard never trips.
The issue body has been left as filed and this plan corrects it.

Couplings flagged

• Agent question card never renders in the channel: channels.spec.ts:500 hard-fails on every main run since #96 #110 / Agent attention & recovery: distinguish liveness from progress and surface actionable exceptions #105. PR feat(desktop): add agent attention and recovery #108 modifies useLiveHomeFeedActions.ts, the exact file
  behind Agent question card never renders in the channel: channels.spec.ts:500 hard-fails on every main run since #96 #110. Whichever lands second must re-verify channels.spec.ts:500.
• feat(desktop): complete Hermes first-class operations in Crew #104 Phase 03 / Agent attention & recovery: distinguish liveness from progress and surface actionable exceptions #105 Slice 2. Both define the 46040 -> 46041/46042 path.
  PR feat(desktop): add agent attention and recovery #108 already implements it.
• Channel-first missions: promote a thread into durable agent work in place #102 Phase 2 / Agent attention & recovery: distinguish liveness from progress and surface actionable exceptions #105. Channel-first missions: promote a thread into durable agent work in place #102 consumes projections Agent attention & recovery: distinguish liveness from progress and surface actionable exceptions #105 defines.
• MessageThreadPanel.tsx sits at 999/1000 lines: the file-size ratchet will break on the next upstream sync #111 / Channel-first missions: promote a thread into durable agent work in place #102. Channel-first missions: promote a thread into durable agent work in place #102 attaches to MessageThreadPanel.tsx, at 999/1000 lines.

Verification

Docs-only change. No tests apply. Links in the index resolve to sibling plan dirs.

[oscarlehuu]

Closing per founder decision (2026-08-10): this sequencing predates the north star lock (#115) and the current issue generation. Superseded by issues #116–#122 and the plans PR for #117/#118/#119/#121. Portions of the six-issue sequencing that remain relevant were re-derived in the new issues; anything still missing should be filed as a new issue against the current docs rather than revived from this branch.