zy95-12 · Nov 4, 2026 · Nov 4, 2026 · Nov 4, 2026
diff --git a/LESSONS.md b/LESSONS.md
@@ -13,6 +13,14 @@ Dated, short lessons from real FlyBot work. Agents read this **before coding**.
 
 ## Open lessons
 
+### 2026-11-04 — Missing workerId does not define admission cleanup (Seen 1×)
+
+Binding can fail before execute returns; register reservation ownership and cleanup
+across preparation/binding/spawn. Persist the spawn fence before external effects.
+Only proven pre-spawn failure permits rollback; retain unknown occupancy and original
+context, and serialize late startup against reclamation.
+
+
 ### 2026-11-04 — An enabled schedule is not a completed inspection (Seen 1×)
 
 A process-local waiter disappears on restart; correlate durable occurrence and Run

diff --git a/PROJECT.md b/PROJECT.md
@@ -8,6 +8,14 @@ architecture lives in [docs/architecture.md](docs/architecture.md) and
 Current stage and priorities: [ROADMAP.md](ROADMAP.md). How decisions are made:
 [docs/governance.md](docs/governance.md). Dated decisions: [decisions/](decisions/README.md).
 
+## Follow-up reliability delivery — 2026-11-04
+
+Owner authorizes sequential #2336/#2337/#2334/#2346/#2338/#2347/#2331/#2342
+implementation, one PR each, tests/memory/docs and validated merge to the delivery
+branch. #2336 establishes owned reservation rollback and a durable pre-spawn fence;
+unknown execution, workspace/context and current authority remain protected.
+No live task/configuration/credential mutation or deployment is implied.
+
 ## Execution recovery — 2026-11-04
 
 #2325 aligns actual reused SDK/resource checkouts with pinned baseline/seed facts,

diff --git a/ROADMAP.md b/ROADMAP.md
@@ -1,5 +1,12 @@
 # FlyBot — Roadmap
 
+## Follow-up reliability — 2026-11-04
+
+Implement #2336 reservation cleanup first; then #2337 CLI configuration, #2334
+environment attribution, #2346 context migration, #2338 receipts, #2347 recovery
+routing, #2331 controlled merge and #2342 operational-agent collaboration. Separate
+PRs retain existing UI/control-plane, authority and exact exit/evidence contracts.
+
 ## Recovery delivery — 2026-11-04
 
 Implement and merge #2325 → #2327 → #2328 → #2330 sequentially into
@@ -52,41 +59,11 @@ Vendor-specific adapters and live external-provider acceptance remain separate.
 Backend delivery does not complete original frontend #41 acceptance
 or merge staged PRs.
 
-S6/#55 integrates current service stages with original native frontend work and
-adds repeatable isolated fault/replay acceptance. Actual PR feedback→original Task→
-Codex Dev→Tester has run; final candidate/Review/Chief evidence and compatible UI
-rollout are recorded separately. Do not replace #41 or claim its whole visual/
-dual-mode matrix from this backend service acceptance.
-
-S5/#54 adds the trusted-private compatibility profile and execution-boundary query,
-without per-Bot OS infrastructure or live identity/sandbox migration. Current-role,
-safe/watch, private-data and concurrency regression suites guard existing rights.
-Exact-head full checks and dedicated-account metadata probes are separate from
-S6/#55 integrated GitHub/SDK/recovery/UI acceptance, which remains required.
-
-S4/#53 implements scoped GitHub feedback→existing Task inputs with persistent
-revisions, bounded source sweeps and native schedule configuration. Full exact-head
-verification, compatible checkpoint rollout and actual GitHub→Bot handoff evidence
-must precede delivery claims. S5/#54 trust/capability boundary and S6/#55 integrated
-fault/recovery acceptance follow. S1–S3/#60 are deployed; original frontend Task40/
-PR48 now advances its new sealed candidate through native Tester/Reviewer roles.
-
-Immediate #60 foreground repair completes approved extra-merge host sealing before
-S2/S3 batch rollout and current frontend Attempt resumption. Keep the existing
-index/MERGE_HEAD, shared-ref restriction, Task/PR and independent candidate gates.
-The first batch is deployed; live source admission found a historical-Chief
-scope regression. PR62 includes its real-state reproduction and focused guard
-repair before resuming the original frontend Attempt. No full frontend acceptance yet.
-
-Service #49 implementation order: S1/#50 (PR56, deployed), S2/#51 (PR58, verified,
-awaiting compatible batch rollout), S3/#52 (PR61,4374 checks and real Codex receipt/pool acceptance verified), then source subscriptions/#53, trusted execution boundary/#54 and
-integrated fault/recovery acceptance/#55. #57/PR59 compatibility repair is deployed;
-original Chief resumed actual frontend Dev without changing Task40/PR48 ownership.
-Each stage keeps memory, architecture and regression evidence with its own PR.
-
-Immediate compatibility repair #57 removes the verified prior-invocation workspace
-and Session resumption blocker before continuing service #51–#55. Preserve the
-current frontend Task/Attempt/PR and unknown-exit protections.
+Earlier service chronology are [preserved in the dated history](docs/history/2026-11-04-earlier-delivery-roadmap.md).
+Current authorization, Task/Session/role constraints and original capability compatibility
+remain in PROJECT.md and canonical subsystem contracts. Historical test evidence does
+not complete outstanding real-provider, mixed-resource, release or frontend acceptance;
+no new deployment, privilege expansion or automatic mission change is authorized.
 
 Current execution order (Owner update, 2026-11-01): keep the existing frontend
 #40 Task/PR48 autonomous while implementing backend #49–#55 directly in the
@@ -98,53 +75,11 @@ live integration, frontend restoration and R8 acceptance retain their existing
 issues; no passing unit suite alone completes them.
 
 
-The owner-approved plan is [fork #1](https://github.com/zy95-12/FlyBot/issues/1), recorded
-in [the fork decision](decisions/2026-10-31-project-first-fork.md). Historical upstream
-stages below do not restrict the fork's approved Project and frontend refactor.
-
-| Milestone | Deliverable | Status / issue |
-| --- | --- | --- |
-| M0 | Isolated DSH/tool/cwd/lifecycle and unchanged-platform HTTP checks | Merged PR #9; [#2](https://github.com/zy95-12/FlyBot/issues/2) |
-| M1 | Project domain and direct Task/Run/Workspace ownership | PR #10 verified locally, awaiting review; [#3](https://github.com/zy95-12/FlyBot/issues/3) |
-| M2 | Shared membership, platform identity adapter and backend authorization | PR #11 verified locally, awaiting review; [#4](https://github.com/zy95-12/FlyBot/issues/4) |
-| M3 | Worktree, code/tests/PR/review evidence and human acceptance | PR #13 locally verified; integrated real-provider acceptance passed in M6; [#5](https://github.com/zy95-12/FlyBot/issues/5) |
-| M4 | Public project Bots and private user/project agents | PR #14 verified locally, awaiting review; [#6](https://github.com/zy95-12/FlyBot/issues/6) |
-| M5 | Rebuild task navigation, Project board, details and resource view | PR #15 locally verified, awaiting review; [#7](https://github.com/zy95-12/FlyBot/issues/7) |
-| M6 | Repeatable two-user end-to-end acceptance | PR #16: real-model/GitHub/browser acceptance passed; awaiting review; [#8](https://github.com/zy95-12/FlyBot/issues/8) |
-
-The owner authorized sequential M1–M6 implementation, one PR per milestone. Each PR
-updates long-term memory, architecture and tests. Merge status is tracked separately.
-
-Reuse platform services unchanged first; new project logic/adapters belong in FlyBot.
-M0's deterministic model and existing shared-write denials do not imply M2/M3 completion.
-GPU pools, multi-host scheduling, auto-merge and exact crash resume remain deferred.
-The 30–43 person-day estimate is provisional pending interfaces/frontend design.
-
-## Authorized M7 sequence
-
-| Phase | Deliverable | State |
-| --- | --- | --- |
-| [M7.1 #21](https://github.com/zy95-12/FlyBot/issues/21) | Durable Chief/role memory and temporary Session snapshots | PR #26; full verify and actual DSH context checks passed |
-| [M7.2 #22](https://github.com/zy95-12/FlyBot/issues/22) | Execution resource registry, candidate pools and leases | PR #27; full verify + actual local/Docker probe; integration #30 |
-| [M7.3 #23](https://github.com/zy95-12/FlyBot/issues/23) | Native/Codex/Claude Agent SDK adapters | Draft PR #28; full verify passed; real authorization acceptance pending |
-| [M7.4 #24](https://github.com/zy95-12/FlyBot/issues/24) | Owner configuration, probes and authorized resource terminal | Draft PR #29; full verify + actual PTY checks; browser integration #30 |
-| [M7.5 #25](https://github.com/zy95-12/FlyBot/issues/25) | Cross-resource code/evidence transfer and real browser acceptance | Draft PR #30; code/resource/browser checks implemented, authenticated mixed-Harness acceptance pending |
-
-User explicitly authorized sequential implementation and a separate PR for every issue,
-including long-term memory, architecture and tests. Merge/deployment is separate.
-
-## Automatic development service sequence
-
-[#49](https://github.com/zy95-12/FlyBot/issues/49) tracks S1 #50 service/TaskView →
-S2 #51 durable Effects/outbox → S3 #52 input/role receipts → S4 #53 external
-Task subscriptions. S5 #54 trusted deployment/authority follows S1 and applies
-throughout; S6 #55 integrated recovery acceptance follows all five. One PR per
-implementation issue; no extra tracking-only PR. S1 PR56 has a deployed initial facade (673c741b); long-history follow-up
-b9cceb02 passed full4333 verification, deployed at a released checkpoint and
-passed three-account scoped HTTP/pagination checks, preserving original UI bytes.
-S2/#51 effect/outbox code and fault regressions are implemented; full verification
-and source-bound rollout follow before S3 inputs/read receipts.
-Independent repository review is not claimed. Original #34–#41 residuals remain.
+Earlier M0–M7 and service checkpoints are [preserved in the dated history](docs/history/2026-11-04-earlier-delivery-roadmap.md).
+Current authorization, Task/Session/role constraints and original capability compatibility
+remain in PROJECT.md and canonical subsystem contracts. Historical test evidence does
+not complete outstanding real-provider, mixed-resource, release or frontend acceptance;
+no new deployment, privilege expansion or automatic mission change is authorized.
 
 ## Historical upstream roadmap
 
@@ -250,48 +185,11 @@ in-process service and incomplete real watch/UI acceptance remain explicit; no
 standalone frontend, OpenAPI service or runtime migration is implied.
 
 
-### Backend convergence #2184 — 2026-11-03
-
-SDK Task-role binding hardening: pinned shared definition snapshots, latest observed
-partial-thread recovery, verified legacy-key adoption, durable dispatch intent,
-explicit incompatible recovery, current rights/resource checks and truthful adapter
-capabilities. Native Task-role context restoration, parent waits, transactional Effects
-and external command receipts remain dependent acceptance; no live-provider or deploy
-claim from controlled process tests. README/API rewrite remains last (#2189).
-
-
-### Collaboration contract convergence — 2026-11-03 (#2175)
-
-Expose durable Task handoffs through one scoped contract projection, with distinct
-pending/accepted/processed/revoked/invalidated states, source/target context references,
-actual claim/worker binding, initiative fences and predecessor/successor history.
-Idempotent trusted offer/accept ports reuse pools; unresolved old writers cannot be
-bypassed by Chief rerouting to another resource. Consultation is a separate private
-protocol over retained transcripts, without Task creation or initiative transfer.
-Chief prefers an active personal agent for user questions: safe answers require user
-publication, watch answers use existing inbox rights; only committed Effects may wake
-an authorized native watch. External clients have isolated mailboxes, not fabricated
-provider sessions. SDK process/HTTP tests and real-provider acceptance remain distinct.
-Native restoration/context bounds, transaction events and parent waits stay on their
-original dependent issues; no control-plane, deployment or frontend changes.
-
-
-### Context/memory convergence — 2026-11-03 (#2177)
-
-Unify model/Harness context budgeting and bounded source tools across persistent
-Chief and temporary roles. Unknown capacity stays explicit; full rules/current
-constraints remain required, retrieval yields before overflow, and native/SDK input
-is checked before delivery. Working summaries use original public role-memory
-records with exact source/evidence/uncertainty, separate resident Chief topics and
-rollback of uncommitted staging. Facts have source/classification/version/archive;
-role-authored experience cannot declare facts or overwrite human control. Private
-notes retain history under bounded views. Codex tool history has an explicit cap;
-SDK compaction remains runtime-managed, without claiming aggregate usage is current
-window pressure. No new memory platform, executor or timer. Native Task-role physical
-restore2184 and normalized SQL storage2182 remain separate acceptance; real SDK
-verification must name exact runtime source hashes, not borrow deterministic tests.
-
-#2177 follow-up: validate two sequential native Chief jobs against fresh per-turn facts while retaining the same resident Session; dependent Task admission runtime acceptance remains #2179.
+Earlier convergence implementation notes are [preserved in the dated history](docs/history/2026-11-04-earlier-delivery-roadmap.md).
+Current authorization, Task/Session/role constraints and original capability compatibility
+remain in PROJECT.md and canonical subsystem contracts. Historical test evidence does
+not complete outstanding real-provider, mixed-resource, release or frontend acceptance;
+no new deployment, privilege expansion or automatic mission change is authorized.
 
 ### Completed convergence checkpoints
 

diff --git a/changelog.d/2336.fixed.md b/changelog.d/2336.fixed.md
@@ -0,0 +1 @@
+Reclaim proven orphan reservations and roll back SDK admission failures before external spawn. Preserve unknown execution capacity, original Task-role context and workspaces with durable ownership/startup evidence.
diff --git a/core/src/host/hooks.js b/core/src/host/hooks.js
@@ -1393,6 +1393,13 @@ export function install(rt) {
             ready({ task, definition }) {
               try { rt.projectResources.plan(task.projectId, { roles: [definition] }); return true } catch { return false }
             },
+            async abort({ task, claim, reason }) {
+              const attempt = rt.state.projectAttempts?.[claim.attemptId]
+              for (const session of Object.values(attempt?.roleSessions ?? {})) if (session.poolClaimId === claim.id && session.botId) {
+                try { await lifecycle.release(session.botId) } catch { /* Do not assume unknown exit. */ }
+              }
+              return rt.projectDelivery.abortPoolAdmission(task, claim, reason)
+            },
             async execute({ task, claim, definition }) { return lifecycle.spawn(await rt.projectDelivery.preparePoolRole(task, claim, definition)) },
             async dispatch({ task, claim, workerId }) {
               const taskInbox = await rt.projectInputs?.prepare(task.id, claim)
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		Reclaim proven orphan reservations and roll back SDK admission failures before external spawn. Preserve unknown execution capacity, original Task-role context and workspaces with durable ownership/startup evidence.