task: Implement Kiro spec pre-oss-core-slimming

[matdev83]

Implement the approved Kiro SDD introduced by merged spec-only PR #557.

Spec: .kiro/specs/pre-oss-core-slimming/

Current implementation status

The production implementation and runtime verification are effectively complete on current main:

• feat(pre-oss): closed-plane contract on generated planes (1.x+2.x) #576 — tasks 1.x + 2.x: corrected baseline and selected sdk(feature): Decide support contract for dynamically declared planes #554 closed generated-plane contract;
• feat(pre-oss): move tool-call repair into feature (3.x) #582 — task 3.x: tool-call repair moved into feature ownership;
• feat(pre-oss): move secret-guard into feature + compose adapter (4.x) #578 — task 4.x: secret-guard implementation moved into feature ownership;
• feat(pre-oss): relocate compaction detector to infra (5.x) #579 — task 5.x: compaction detector moved to infra behind the frozen consumer port;
• feat(pre-oss): runtimebundle adapters + zero feature imports (6.x) #580 — task 6.x: runtimebundle adapters and zero direct concrete-feature imports;
• feat(pre-oss): ownership ratchets, SDK fixture, docs, evidence (7.x+8.x) #581 — tasks 7.x + 8.1–8.3: ownership ratchets, external SDK fixture/docs, performance evidence and residual ownership inventory;
• chore(pre-oss): pin inventory SHAs to merged main #584/fix(pre-oss): baseline SHA validated as mainline ancestor #585/fix(pre-oss): hermetic git identity in inventory SHA test #586 — merged-main fixes for residual-inventory SHA identity/ancestor verification;
• ci: add pre-OSS exact race gate #588 — exact SHA-addressable Linux race workflow;
• fix(runtime): synchronize concurrent terminal state #589/test(runtime): isolate parallel stream assembly state #590 — runtime/test race fixes discovered by that exact gate;
• fix(feature): make generated plane policy canonical #591 — canonical generated-plane policy correction and final production correction pass at d784a8344888dd9de2141a13d4bf723125d4b08c.

Authoritative Linux race run 33891386913 executed the exact required package scope against #591's merged main SHA and completed successfully. The #571 closeout comment also records generator/check, SDK/featurebundle/archtest/speccheck, external fixture, runtime-sensitive/migrated-feature suites, PR CI/QA, and benchmark allocation/B-op evidence as green.

Remaining closeout gate

This work order is not yet truthfully complete in repository state. On current main:

• .kiro/specs/pre-oss-core-slimming/tasks.md still leaves Task 8.4 unchecked;
• .kiro/specs/pre-oss-core-slimming/spec.json still has phase: tasks-generated, ready_for_implementation: true, and completed: false;
• the spec remains under the active .kiro/specs/ location rather than being archived as Task 8.4 requires.

Therefore the earlier completed-state closure was premature. The remaining work is final Kiro/repository closeout: reconcile Task 8.4 with the already-green merged-main evidence, ensure the required independent architecture/code-review disposition is recorded, archive the spec while retaining the residual ownership inventory, and mark the canonical spec completed.

Do not reopen production scope merely to satisfy bookkeeping; only repair missing evidence/closeout required by Task 8.4. Close #571 once those canonical artifacts truthfully represent the completed implementation.

Tracker relationship

• This is the implementation work order for merged spec PR spec: define pre-OSS core slimming and feature ownership closure #557.
• Directly blocks feat(split): OSS / Enterprise split [ENDS OSS Base Milestone] ============================================ #398 because spec: define pre-OSS core slimming and feature ownership closure #557 defines this work as release-bounded pre-OSS closure and there is no separate feature-level parent tracker for the overall program.
• sdk(feature): Decide support contract for dynamically declared planes #554 is the contained compatibility subtracker. Its production contract and required merged-main/Linux-race evidence are now present, so its completed state may remain closed; do not duplicate sdk(feature): Decide support contract for dynamically declared planes #554 directly onto feat(split): OSS / Enterprise split [ENDS OSS Base Milestone] ============================================ #398.
• Preserve the terminal relationship as task: Implement Kiro spec pre-oss-core-slimming #571 → feat(split): OSS / Enterprise split [ENDS OSS Base Milestone] ============================================ #398.

Dependencies / coordination

• No active hard prerequisite remains inside OSS Base.
• audit(performance): Reality check for high volume, highly concurrent proxy use #394 performance hardening remains independent; this task's same-host microbench evidence is not audit(performance): Reality check for high volume, highly concurrent proxy use #394 load/HOLD certification.
• task: Implement Kiro spec large-payload-streaming-fast-path #532/feat(performance): Add large-payload streaming fast path to avoid full request materialization #503 large-payload work may proceed independently against current post-slimming ownership.
• task: Implement Kiro spec bulk-inference-provider-expansion #545 bulk-provider expansion is independent.
• feat(rebranding): Complete project rebranding #429 rebranding is coordination only, not a blocker.

Follow-up boundary

Merged spec PR #559 / post-split task #572 is the second/final architecture-completion phase. Its Task 0 requires an implemented and certified predecessor. Production implementation and exact race evidence are present, but canonical pre-OSS completion/archive is not yet recorded. Keep #572 planned until this closeout is complete.

Because #572 is in the later milestone, do not model #571 → #572 as a GitHub dependency edge; milestone sequencing supplies that prerequisite.

[matdev83]

Corrective continuation completed and merged via PR #591 at d784a83. Generator/check, SDK/featurebundle/archtest/speccheck, external fixture, runtime-sensitive packages, migrated feature suites, PR CI/QA, and exact Linux race run 33891386913 are green. Benchmark allocations and B/op remain unchanged across the 31-case extension-plane family; timing samples were noisy but introduced no allocation regression. No core-feature-ownership-full-closure (#572) work was pulled forward.