traverse-framework · enricopiovesan · Sep 6, 2026 · Sep 6, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -32,20 +32,20 @@ gh project item-list 2 --owner traverse-framework --format json --limit 300 \
 | `new-app-author-e2e` | **Done** (#275) | CLI `app new` → App-Refs layout + Web host rewrite |
 | `fix-nightly-native-macos-windows` | **Done** (#277) | Swift 6 Sendable mappers + WinUI test compile of PresentationMapper |
 | `llm-mcp-mode-a-spec119-scaffold` | **Done** (#271) | Spec 119 consumer scaffold (fail-closed); live kit execute stays `llm-mcp-traverse-starter-catalog` |
-| `llm-mcp-traverse-starter-catalog` | **Blocked** | Live kit execute blocked on Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241) (Spec 119 Mode A implement) |
+| `llm-mcp-traverse-starter-catalog` | **Blocked** | Live kit execute blocked on Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241) (Spec 119 Mode A); absorbed Mode A cutover from `llm-mcp-0-10-live-cutover` |
 | `loop-wf1-registry-deps` | **Done** (#263) | Digest inventory in `docs/loop-registry-deps.md` |
 | `loop-wf1-multi-os` | **Done** (#265) | `apps/loop/` WF1 × 7 OS; compose via `registry_ref` |
 | `retire-registry-ref-materialize` | **Done** (#267) | Phase A gate; host cutover → Future `retire-registry-ref-materialize-hosts` |
 | `retire-registry-ref-materialize-hosts` | **Future** | Full materialize delete after BundleEmbedder + Spec 107 |
 | `meeting-notes-remaining-natives` | **Done** (#236) | Shipped via `meeting-notes-wave2-os-ports` |
-| `llm-mcp-embedded-host` | **Blocked** | Scaffold Done (#268); live Mode B blocked on Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242) |
+| `llm-mcp-embedded-host` | **Future** | Scaffold Done (#268); Mode B host CLI on Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242) — not active until Mode A unlock progresses |
 | `two-app-reuse-contract` | **Done** (#285) | Pin `meeting-notes.process` 1.3.2 for meeting-notes + loop |
 | `two-app-reuse-execute` | **Done** (#286) | Align pins + wasmtime evidence |
 | `two-app-reuse-execute-ci` | **Done** (#287) | Fixture-only evidence (no Traverse core examples) |
 | `pin-ci-traverse-0-10` | **Done** (#292) | CI / embedder / runtime.wasm → Traverse v0.10.0 |
-| `llm-mcp-0-10-live-cutover` | **Blocked** | v0.10.0 Mode A still expedition-only — Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241); Mode B host [#1242](https://github.com/traverse-framework/Traverse/issues/1242) |
+| `llm-mcp-0-10-live-cutover` | **Done** (absorbed) | Mode A → `llm-mcp-traverse-starter-catalog`; Mode B → `llm-mcp-embedded-host` Future; pin already #292 |
 | `two-app-reuse-host-execute` | **Blocked** | Traverse [#1240](https://github.com/traverse-framework/traverse/issues/1240) — BundleEmbedder still fails WASI `proc_exit(0)` on 1.3.2 |
-| `two-app-reuse-lifecycle` | **Blocked** | After `#282` — upgrade/deprecation outcomes |
+| `two-app-reuse-lifecycle` | **Ready** | Fixture pin-flip + documented deprecation for meeting-notes + loop (not BundleEmbedder; not production release-train) |
 
 Full gap table + wave notes: [`docs/production-reference-plan.md`](docs/production-reference-plan.md).
 

diff --git a/apps/llm-mcp-reference/README.md b/apps/llm-mcp-reference/README.md
@@ -51,7 +51,7 @@ Stronger isolation path: prepare registry deps into a **host-owned verified cach
 
 See [`mode-b/README.md`](mode-b/README.md), [`mode-b/mcp.json.example`](mode-b/mcp.json.example), and fail-closed [`mode-b/serve.sh`](mode-b/serve.sh).
 
-**Honest status:** Spec 520 library APIs shipped in Traverse (#860). The Mode B MCP host CLI is **not** shipped (Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242)). Expedition Mode A remains the only runnable LLM path on v0.10.0; Mode B launcher exits non-zero until that host exists.
+**Honest status:** Spec 520 library APIs shipped in Traverse (#860). The Mode B MCP host CLI is **not** shipped (Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242)). App-Refs ticket `llm-mcp-embedded-host` is **Future** (Mode A kit path is the active unlock). Expedition Mode A remains the only runnable LLM path on v0.10.0; Mode B launcher exits non-zero until that host exists.
 
 Optional bearer token (execution commands):
 

diff --git a/docs/adr/0006-two-app-reuse-contract.md b/docs/adr/0006-two-app-reuse-contract.md
@@ -13,7 +13,7 @@ Traverse #1168 requires proof that a published capability is a reusable platform
 
 Pin **both** `meeting-notes` and `loop` to public `meeting-notes.process` **1.3.2** (`registry_ref.version_range` exactly `1.3.2`, artifact digest `sha256:ec192a0c2104b08bee76418c5c6d44358036568d655d8465e506858d1aaadbf2`). Full pin, host routes, and current drift: [`../two-app-reuse-contract.md`](../two-app-reuse-contract.md).
 
-This ADR does not implement alignment or execution. `#282` applies the pin and produces evidence. `#283` covers upgrade/deprecation.
+This ADR does not implement alignment or execution. `#286` applies the pin and produces evidence. `#283` (`two-app-reuse-lifecycle`) covers fixture-level upgrade/deprecation outcomes (Ready — see decision-log 2026-09-05).
 
 ## Consequences
 

diff --git a/docs/decision-log.md b/docs/decision-log.md
@@ -384,3 +384,127 @@ Append-only record of design decisions for App-References. Newest sessions at th
 **Decision:** Exact pin `1.3.2` / digest `sha256:ec192a0c…` for both apps. Record in [`two-app-reuse-contract.md`](two-app-reuse-contract.md) and ADR [`0006`](adr/0006-two-app-reuse-contract.md). No manifest alignment in this ticket.
 
 **Why:** Same capability already sits on two `app_id` boundaries with different user-facing purposes. Floating `^` ranges are exactly what #1168 forbids.
+
+---
+
+## 2026-09-05 — Blocked tickets ownership walk
+
+**Context:** Brainstorm over Project 2 Blocked tickets to classify ownership (on user vs Traverse vs defer) and decide board hygiene vs upstream push order. Live board query was intermittently rate-limited; inventory from `AGENTS.md` + open Traverse issues #1240/#1241/#1242.
+
+### Brainstorm goal
+
+**Question:** What should this brainstorm decide?
+
+**Options considered:**
+- Prioritize which Traverse blocker to push next — pros: focuses unlock; cons: skips hygiene
+- Decide App-Refs interim workarounds while Traverse is blocked — pros: keeps shipping; cons: fake-runtime risk
+- Board hygiene only (keep / Future / narrow DoD) — pros: cleans false urgency; cons: no upstream move
+- Walk every blocked ticket and classify ownership — pros: full map; collapses shared root causes; cons: longer
+
+**Recommendation:** Ownership walk first.
+
+**Decision:** Walk every blocked ticket and classify (on you / on Traverse / defer).
+
+**Why:** Three of five share two Traverse issues; a pass usually collapses the set before picking priority or workarounds.
+
+### LLM MCP cluster
+
+**Question:** How to classify `llm-mcp-traverse-starter-catalog`, `llm-mcp-embedded-host`, and `llm-mcp-0-10-live-cutover`?
+
+**Options considered:**
+- On Traverse only; leave all Blocked — pros: honest; cons: no prioritization signal
+- On user to prioritize/assign Traverse #1241/#1242 — pros: names unlock; cons: needs Traverse ownership
+- Defer whole cluster to Future — pros: less Blocked noise; cons: hides Approved Spec 119 wait
+- Split: Mode A (#1241) active Blocked; Mode B (#1242) → Future — pros: Mode A unblocks kit path first; cons: cutover umbrella spans both
+
+**Recommendation:** Split Mode A / Mode B.
+
+**Decision:** Mode A stays active Blocked on Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241); Mode B (`llm-mcp-embedded-host`) → Future on [#1242](https://github.com/traverse-framework/Traverse/issues/1242).
+
+**Why:** Mode A is the Cursor/Claude kit path; Mode B is a separate host product.
+
+### Cutover umbrella
+
+**Question:** What happens to `llm-mcp-0-10-live-cutover` after the Mode A/B split?
+
+**Options considered:**
+- Narrow DoD to Mode A only; stay Blocked — pros: simple; cons: Mode B cutover needs a home
+- Split into Mode A + Mode B cutover tickets — pros: clean; cons: overlaps live tickets
+- Keep umbrella Blocked on both — pros: no rewrite; cons: Mode B blocks Mode A Done
+- Cancel/absorb into `llm-mcp-traverse-starter-catalog` + `llm-mcp-embedded-host` — pros: fewest tickets; cons: lose explicit post-pin checklist unless folded into those DoDs
+
+**Recommendation:** Absorb/cancel umbrella.
+
+**Decision:** Cancel/absorb `llm-mcp-0-10-live-cutover` into the Mode A catalog and Mode B embedded-host tickets (pin already Done via `pin-ci-traverse-0-10`).
+
+**Why:** After the pin, the umbrella duplicates the two live tickets.
+
+### Two-app host execute
+
+**Question:** How to classify `two-app-reuse-host-execute` (Traverse [#1240](https://github.com/traverse-framework/Traverse/issues/1240))?
+
+**Options considered:**
+- On Traverse only; leave Blocked — pros: no fake host path; cons: public BundleEmbedder proof incomplete
+- On user to prioritize #1240 — pros: names unlock; cons: needs Traverse triage ownership
+- Downgrade to Future with materialize-retire wave — pros: less noise; cons: understates open bug
+- Accept wasmtime-only as Done; Future/Cancel host-execute — pros: paper-unblocks reuse; cons: weakens public-host proof
+
+**Recommendation:** Stay honestly Blocked on #1240.
+
+**Decision:** Leave `two-app-reuse-host-execute` Blocked on Traverse #1240; do not redefine Done around wasmtime alone.
+
+**Why:** Ticket exists for public BundleEmbedder proof; App-Refs must not invent substitute WASM.
+
+### Lifecycle predecessor
+
+**Question:** What to do with `two-app-reuse-lifecycle` (board note still “After `#282`”)?
+
+**Options considered:**
+- Move toward Ready (execute Done is enough) — pros: clears false Blocked; cons: may need registry behaviors
+- Retarget Blocked to host-execute / #1240 — pros: if DoD needs BundleEmbedder; cons: couples unrelated concerns
+- Future until a real published upgrade/deprecation — pros: live release train; cons: long wait
+- Blocked on user for Spec/DoD rewrite first — pros: avoids wrong proof; cons: design beat before Ready
+
+**Recommendation:** Ready path if DoD is pin/upgrade semantics without BundleEmbedder.
+
+**Decision:** Not blocked on `#282` (execute already Done as #286) or on #1240; proceed to Ready after DoD is made checkable.
+
+**Why:** Written scope is upgrade/deprecation outcomes for the shared pin pair; no BundleEmbedder requirement in the contract table.
+
+### Lifecycle Spec/DoD shape
+
+**Question:** Before Ready, how to make lifecycle DoD claimable?
+
+**Options considered:**
+- Ready as-is; flesh DoD in implementing PR — pros: fastest; cons: weak claim gate
+- Full Spec/DoD rewrite brainstorm first — pros: strongest gate; cons: extra design pass
+- Future until registry publishes next `meeting-notes.process` — pros: real train; cons: App-Refs can’t drive calendar
+- Fixture-only pin flip + documented deprecation handling (no new registry release) — pros: claimable soon; honest; cons: weaker than production upgrade proof
+
+**Recommendation:** Fixture-level DoD, then Ready.
+
+**Decision:** Define DoD around fixture-only pin flip + documented deprecation handling (explicit non-claim of production release-train proof), then set status Ready.
+
+**Why:** Keeps the ticket honest and claimable without waiting on a registry publish or BundleEmbedder.
+
+### Next action order
+
+**Question:** With two honest Blocked tickets left, what to push first?
+
+**Options considered:**
+- Prioritize Traverse #1241 (Mode A kit MCP) — pros: live kit façades; cons: host-execute stays stuck
+- Prioritize Traverse #1240 (BundleEmbedder) — pros: completes #1168 host proof; cons: MCP stays expedition-only
+- App-Refs board hygiene first; Traverse priority later — pros: lifecycle becomes claimable now; cons: upstream idle briefly
+- Parallel triage of #1241 + #1240 plus hygiene — pros: no false sequencing; cons: split attention
+
+**Recommendation:** Hygiene first, then #1241.
+
+**Decision:** Apply App-Refs board hygiene first (Mode B → Future; cancel cutover umbrella; Ready lifecycle with fixture DoD; refresh `AGENTS.md` / README blockers). Then push Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241) as the higher-leverage remaining Blocked unlock. Leave #1240 Blocked without redefining it.
+
+**Why:** Hygiene unlocks Ready work without waiting on Traverse; Mode A MCP is the stronger demo unlock of the two remaining upstream bugs.
+
+### What was explicitly deferred
+
+- Assigning/staffing Traverse #1240 vs #1241 beyond “#1241 after hygiene”
+- Any App-Refs interim workaround that fakes kit MCP or BundleEmbedder success
+- Production registry release-train proof for lifecycle (fixture-only by design)
diff --git a/docs/llm-reference-apps-plan.md b/docs/llm-reference-apps-plan.md
@@ -76,7 +76,7 @@ Shared:
 | Mode | When | Notes |
 |---|---|---|
 | **A. MCP stdio → local Traverse** (v1 default) | Developer laptop / agent IDE | Expedition bootstrap only on Traverse **v0.10.0** (`cargo run -p traverse-mcp -- stdio`). Spec 119 public-registry host: **scaffolded** under `apps/llm-mcp-reference/mode-a/` (fail-closed). Live kit discover/execute blocked on Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241) |
-| **B. MCP → embedded host in a sidecar process** | Stronger product isolation | **Scaffolded** under `apps/llm-mcp-reference/mode-b/` (Spec 520 prepare/cache docs + fail-closed launcher). Mode B host CLI **not shipped** — Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242) |
+| **B. MCP → embedded host in a sidecar process** | Stronger product isolation | **Scaffolded** under `apps/llm-mcp-reference/mode-b/` (Spec 520 prepare/cache docs + fail-closed launcher). Mode B host CLI **not shipped** — Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242). App-Refs ticket `llm-mcp-embedded-host` is **Future** (Mode A is the active unlock) |
 | **C. Remote MCP gateway** | Multi-tenant SaaS | Future; needs auth/tenancy — not this slice |
 
 v1 documents **Mode A** only. Do not revive HTTP `traverse-cli serve` as the production architecture for primary OS shells; MCP stdio is a **separate agent façade**, not a replacement for embedded Web/iOS/Android clients.
@@ -93,8 +93,8 @@ v1 documents **Mode A** only. Do not revive HTTP `traverse-cli serve` as the pro
 | `llm-mcp-grok-adapter` | Grok tool-calling mapping | Future / Ready when API stable |
 | `llm-mcp-mode-a-spec119-scaffold` | Spec 119 Mode A consumer (fail-closed serve + prepare + MCP example) | Done (#271) |
 | `llm-mcp-traverse-starter-catalog` | Discover/execute kit capabilities (`traverse-starter.*` / meeting-notes) via Mode A public registry discovery | Blocked — Spec [`119-verified-registry-mcp-mode-a`](https://github.com/traverse-framework/Traverse/blob/main/specs/119-verified-registry-mcp-mode-a/spec.md) **Approved**; implement Traverse [#1241](https://github.com/traverse-framework/Traverse/issues/1241). v0.10.0 remains expedition-only. First release has no kit content groups (FR-007). Registry [#99](https://github.com/traverse-framework/registry/issues/99) closed |
-| `llm-mcp-embedded-host` | Mode B embedded prepare/cache for MCP host | Scaffold Done (#268); live Mode B host blocked on Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242) (Spec 520 host CLI; Mode B out of Mode A spec #1125) |
-| `llm-mcp-0-10-live-cutover` | Re-verify Mode A kit execute + Mode B host after Traverse v0.10.0 pin | **Blocked** — pin Done (#292); Mode A kit execute missing on tag → [#1241](https://github.com/traverse-framework/Traverse/issues/1241); Mode B host → [#1242](https://github.com/traverse-framework/Traverse/issues/1242) |
+| `llm-mcp-embedded-host` | Mode B embedded prepare/cache for MCP host | Scaffold Done (#268); **Future** until Traverse [#1242](https://github.com/traverse-framework/Traverse/issues/1242) (Spec 520 host CLI). Absorbed Mode B slice of cancelled umbrella `llm-mcp-0-10-live-cutover` |
+| `llm-mcp-0-10-live-cutover` | Re-verify Mode A kit execute + Mode B host after Traverse v0.10.0 pin | **Done (absorbed)** — Mode A live proof stays on `llm-mcp-traverse-starter-catalog` (#1241); Mode B on `llm-mcp-embedded-host` Future (#1242); pin Done (#292) |
 
 ## Success criteria (plan slice)
 

diff --git a/docs/two-app-reuse-contract.md b/docs/two-app-reuse-contract.md
@@ -97,8 +97,10 @@ Host CLI execute via public `BundleEmbedder` (`meeting-notes-cli` / `loop-cli`)
 
 | Ticket | Role |
 |---|---|
-| `#282` `two-app-reuse-execute` | Align meeting-notes pin; execute both apps; publish digest-equal evidence |
-| `#283` `two-app-reuse-lifecycle` | Compatible upgrade + deprecation outcomes for the same pair |
+| `#286` `two-app-reuse-execute` | Align meeting-notes pin; execute both apps; publish digest-equal evidence (Done) |
+| `#287` `two-app-reuse-execute-ci` | Fixture-only CI evidence (Done) |
+| `two-app-reuse-host-execute` | Public BundleEmbedder CLI execute — Blocked on Traverse #1240 |
+| `#283` `two-app-reuse-lifecycle` | **Ready** — fixture pin-flip + documented deprecation for the same pair (not BundleEmbedder; not production release-train). See `docs/decision-log.md` 2026-09-05 |
 
 ## Re-verification