Promote capacity execution verification notes through staging into main

docs/agent-architecture.md

@@ -550,6 +550,8 @@ Stop and record a blocker if exact authority, isolation, project access, provide
 
 ### Current active blocker
 
+- October 6 continuation: capacity-provider and Agent execution test resolution and implementation alignment are active; UI work is deferred. [Agent #321](https://github.com/treeseed-ai/agent/issues/321) records staging `eef952e` (PR #332): complete 840-test verification plus a fresh complete 840-test prerequisite suite before 586 passing coded component steps. Main promotion requires its own fresh CI. [API #494](https://github.com/treeseed-ai/api/issues/494) records 1,915 passing native control-plane tests across 367 complete files on held source, but strict compilation remains failing; later changes need fresh verification. New upstream-error regressions preserve genuine failures and exact retries. [Platform #626](https://github.com/treeseed-ai/platform/issues/626) tracks lineage-preserving staging-to-main cleanup. These are scoped component results, not complete semantic coverage, managed physical closure or an SDK golden pass. The dated observations below remain historical evidence, not instructions to restart the earlier authoring-only gate.
+
 - Human-approved priority clarification (October 4): the governed Proposal work item owns optional safe integer priority, projected unchanged onto its Actor and Reviewer nodes; higher values rank first only after eligibility and project/class fairness, and omission means zero. SDK validation/selection, native API producer/persistence/admission, and managed historical selection replay tests are authored in [API #490](https://github.com/treeseed-ai/api/issues/490), but remain unexecuted. This target clarification is not an implementation or passing acceptance claim.
 
 - Architecture-first test correction (October 2): [Agent312](https://github.com/treeseed-ai/agent/issues/312) now tests profile-owned instructions, class-rename invariance, exact-grant output selection, explicit integration, ordinary deterministic-handler admission/closure, immutable grants/paths/deadlines, and identity-neutral observed verification through native Node/Kernel/Git/runner boundaries. Focused regressions reproduce violations; component scene bindings are declared, not executed acceptance. Per the direct user instruction, unrelated testing and Agent319 delivery are held while these architectural gates are corrected. Full architecture/assignment acceptance remains open; the older observations below are historical, not the current frontier.

docs/agent-assignments.md

@@ -439,6 +439,8 @@ This plan adds only assignment-specific responsibilities to the ownership table
 
 ### Current active blocker
 
+- October 6 continuation: implementation alignment follows the capacity-provider and Agent execution regressions; UI work remains deferred. [Agent #321](https://github.com/treeseed-ai/agent/issues/321) records PR #332 merged to staging `eef952e` after complete 840-test verification and a new complete 840-test prerequisite suite before 586 passing coded component steps; main promotion still needs fresh exact-candidate verification. [API #494](https://github.com/treeseed-ai/api/issues/494) records 1,915 passing tests in 367 complete native control-plane files on unchanged source, with strict compilation still blocking delivery. New upstream-error tests retain failed observations and unchanged-input retry evidence. [Platform #626](https://github.com/treeseed-ai/platform/issues/626) tracks history-preserving cleanup through staging to main, without new tags. Whole assignment/graph semantics, native all-attempt usage and physical closure, and the full SDK golden remain unproven. The older dated observations below are retained history, not the current execution gate.
+
 - Architecture-first test correction (October 2): [Agent312](https://github.com/treeseed-ai/agent/issues/312) reproduces handler-side grant/path widening and deadline mutation through real isolated Git publications, implicit integration guidance outside an explicit integration assignment, and name-dependent handling of native failed verification. The original assignment input remains unchanged, which does not prove immutable authority inside the Kernel. The architecture plan records the reopened enforcement gate. These focused failures block further unrelated testing; proposal/graph/review/settlement and full CLI acceptance remain separately unproven. Preserve the historical DQ/DR evidence below; it is not the current frontier.
 
 - API lifecycle unit and real SQL tests now require YAML-owned activity dependencies and identity-neutral reporting, with exactly one selected workday closeout Reporter. Six focused cases fail. Existing graph guarantees now require the living team graph, exact content/decision/review authority and ordinary weighted admission, not retired group/signal graphs, deliverable manifests or capacity tiers. Their full executable acceptance coverage remains open. Pure test-first development is required after all three layers are aligned; no expectation may be changed merely to fit implementation.