Orchestrata Ultimate › Plans › worktree-merge-plan
Worktree merge plan — 2026-07-29 (session 256)
Source: plans/worktree-merge-plan.md at 8a2cb42ad · last changed 2026-07-29 · first committed 2026-07-29 · 75 lines.
Prose pending.
Plan text
Worktree merge plan — 2026-07-29 (session 256)
Ruling applied: only complete-and-merged branches get deleted; unmerged work gets a completeness-ranked merge plan. Stubs are forbidden — both surviving branches were audited: zero stub/TODO/NotImplemented markers; the single debt marker found is called out below as a blocking item.
Deleted (complete + merged + clean)
| Worktree | Branch | Evidence |
|---|---|---|
| wt-B1 | task/B1 | 0 unmerged commits, clean tree |
| wt-B2 | task/B2 | 0 unmerged commits, clean tree |
| wt-ENGINE | feat/rename-engine | 0 unmerged commits, clean tree |
| wt-FF | fix/fireflies-teardown | 0 unmerged commits, clean tree |
1. feat/hub-link (wt-HUB) — completeness HIGH, merge FIRST
8 commits, +4,677/−150, clean tree. Design plan + implementation briefs are
in-branch (RUL-022 satisfied), implementation spans
knowledge_graph/kernel/gateway/session/orchestrator, includes
scripts/repair_orphan_hub_links.py with an 874-line test module, and the
final commit closes DoD coverage gaps (fact_graph 100%, repair script 99%).
This is the KG root-connectivity work (session-254 lineage — pairs with
check_kg_connectivity.py and the multi-project invariant).
Gap: 388 commits behind master, and the graph-rename merge touched the same knowledge_graph surface — semantic conflicts are guaranteed, not hypothetical.
Steps:
- In wt-HUB:
git merge masterinto feat/hub-link; resolve conflicts against specs (RUL-013), expect them concentrated in knowledge_graph/sync + edge constants. - Run
scripts/check_kg_connectivity.py+ the branch's own test modules, then the full authoritative gate (scripts/run_tests.py, kernel verified). - RUL-008 reviews (code/test/architecture + deterministic gates).
- Merge to master ONLY on 0 failed / 0 errors / 0 skipped; push; delete worktree + branch (RUL-021).
Estimate: one focused session.
2. feat/orchestrata-cli (wt-QWEN) — completeness LOW-MED, build-out needed
The branch has no unique commits (it sat at the old master tip); the
work lives as worktree WIP: +586/−65 in kernel/cli_main.py plus a new
kernel/cli_client.py — the SPEC-000 thin-client CLI over the gateway's
WebSocket JSON-RPC surface (init/status/team/intent commands, .env walk-up,
short-lived JWT minting, sync gateway_rpc helper). Checkpoint-committed
onto the branch as part of this plan so it can no longer be lost.
Blocking items before merge (in order):
- Delete the
# mypy: ignore-errorsheader in cli_client.py and make the module type-clean — the only debt marker found in either branch. - Tests: NONE exist for the new surface. Write real-infra tests per RUL-002 (no mocks): CLI → the RUNNING gateway (auth round-trip, status, team add/remove, intent submit); gateway-down paths fail loudly, never skip.
- Wiring per RUL-012/RUL-023: the
orchestrataCLI entrypoint must be registered (pyproject[project.scripts]) and documented — an unregistered CLI is not a delivery. - RUL-008 reviews + full gate; merge only fully green; delete worktree + branch.
Estimate: 1–2 sessions; the real-infra test surface is the bulk.
Decision for Mohamed: this was kimi/qwen-delegated work (IRC fleet, RUL-018 lane). Reassign the build-out to a builder subagent, or route it back through the IRC task queue?
Order
hub-link first (near-done, unblocks KG-connectivity value adjacent to the P1 "KG schema" backlog item), then orchestrata-cli.