feat: populate the solution-simplicity and code-simplicity strata

CHANGELOG.md

@@ -4,11 +4,33 @@ summary: Chronological history of repository and skill changes.
 
 # Changelog
 
-## 2026-07-27 — Enforced acceptance-gated closeout, populated the correctness stratum, and recovered carved suffixes
+## 2026-07-27 — Populated the solution-simplicity and code-simplicity strata, enforced acceptance-gated closeout, populated the correctness stratum, and recovered carved suffixes
 
+- fix: correct a stale reference, a stale validation entry, and an inverted case
+- fix: make the last two before-state and sanitization defects actually resolved
+  (`41de65daadc5d53bfbb299cb4ecd6d040ac47ab9`)
+- fix: sanitize the repository-history case and correct the changelog order
+  (`3d9fe4925c8908a311453c87ae740bfcf4de20bd`)
+- fix: reconcile records after folding s2 and s3 into one delivery
+  (`5070cf1bbea438e74149dfe0cf9b171a6f7cdb92`)
+- docs: record the code-simplicity delivery and close out corpus population
+  (`875091c32301eafd807d2d5a3e2b402e7ffaca53`)
+- feat: populate the code-simplicity stratum with four adjudicated cases
+  (`f3c064a7bbaf3f89f7a6a5495846b254a54e9a0b`)
+- fix: sweep sanitization across every reviewer-visible field, not only the diff
+  (`ab3921a904ec7835bc4d03ed40b7c8a28d12d2c1`)
+- fix: make every s2 packet internally consistent after the sanitization rename
+  (`1ec231bb17cd0c1db82258756aa7a78e8e7f63ab`)
+- fix: sanitize the solution-simplicity cases against source-vocabulary leakage
+  (`2b56c022c91a925b574a5748112e63bdcbbbf8f2`)
+- docs: record the solution-simplicity delivery and settle the grading method
+  (`da8f53b06072ba0380d01ce06fc4f4a324a6219e`)
+- feat: populate the solution-simplicity stratum with four adjudicated cases
+  (`3105b8e84da78c691f4f93883f39887ff9ae784f`)
 - feat: require acceptance evidence for workflow closeout
+  (`a3597c25ee2d76135d1f0c8642a620e673fc8e57`)
 - fix: make every packet diff a valid patch, and gate the adjudication record
-  (`c4e01d3c6fc94f1bd174d1ad5bd2374511c85471`)
+  (`06a5679643a0a5bcb1944c8bff4bd4986f4f77e1`)
 - fix: stop a grader formulation being quotable from its own packet
   (`fa772a7d770bd3d07f3fdd9bdc45a0c237b1d14e`)
 - docs: record the batch-2 delivery, the clean-control standard, and its limits

review-suite/evals/baseline/v1/ADJUDICATION-PLAN.md

@@ -191,6 +191,40 @@ the opposite of what the table assumed:
 The clean-control question the table flagged as the weakest slot is now settled
 by the owner, and the corpus is built to that standard.
 
+## 2c. Outcome: `s2-solution-simplicity-lens`, no oracle available
+
+Every case in this stratum records `adjudication.second: owner_required`, and a
+test now enforces that a case with no shipped oracle may not record anything
+other than that value. Each case's provenance carries a recommended disposition
+and the specific counter-argument that could overturn it, so the owner is
+confirming or correcting a stated argument rather than starting from a bare case
+description.
+
+| case                             | recommended | the strongest counter-argument, for the owner to weigh                                                                                                                                                                                                              |
+| -------------------------------- | ----------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `setup-service-path-gateway`     | MATERIAL    | A reviewer could value the injected gateway as forward consistency with a future second implementation the packet does not rule out.                                                                                                                                |
+| `registry-client-layering`       | MATERIAL    | The diff only touches the queue-message client; a reviewer could argue the mutation-client duplication predates this change and is out of scope for it.                                                                                                             |
+| `reconciliation-outcome-type`    | CLEAN       | A reviewer could reasonably propose a smaller fix (a second boolean, or one sentinel) instead of a four-member enum - whether that is "over-shoots the requirement" or "unnecessary machinery" is a distinction this case's own accepted non-finding does not draw. |
+| `record-status-transition-guard` | CLEAN       | Retry count and backoff are unspecified in the packet; a reviewer could reasonably ask for that detail without the request being about over-engineering at all.                                                                                                     |
+
+None of these four is settled. They are recommendations, not adjudications, and
+the record is explicit about the distinction so a recommendation is never
+mistaken for a second party's judgement.
+
+## 2d. Outcome: `s3-code-simplicity-lens`, no oracle available
+
+Same shape as `s2`: every case records `adjudication.second: owner_required`,
+enforced by the same test.
+
+| case                                   | recommended | the strongest counter-argument, for the owner to weigh                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        |
+| -------------------------------------- | ----------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `watcher-check-policy-duplication`     | MATERIAL    | Severity is recorded as strong_recommendation rather than blocking, since the packet shows no currently observed drift, only a stated history of it; a reviewer could argue for blocking given that history.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |
+| `metrics-label-formatting-duplication` | MATERIAL    | This case and s2's `registry-client-layering` trace to the same source PR and comment, read at two different levels; the owner should confirm this reuse is deliberate and recorded rather than an unintended duplicate sample of one finding across two strata.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              |
+| `compat-accessor-boundary-duplication` | CLEAN       | A reviewer could reasonably ask whether the two independently scheduled callers could still share one function with two call sites rather than two functions with one body each - a smaller version of the question this case's own accepted non-finding already tolerates.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   |
+| `env-inventory-bullet-format`          | CLEAN       | **Weakest case in this batch.** A reviewer could reasonably propose a narrower fix (wrapping only the wide cell, or a two-column table) instead of the fully verbose bullet form. More importantly: this diff is pure Markdown, and `review-code-simplicity`'s own rubric instructs a reviewer to omit formatting concerns unconditionally. A compliant reviewer therefore says clean regardless of whether the verbosity is justified, which means this case mostly tests rubric compliance rather than the near-miss judgement it was built to demonstrate. The owner should weigh whether to keep it as-is, treat its clean verdict as validating the omission rule rather than a discrimination test, or substitute a code-level (non-Markdown) non-material near-miss in a future batch. |
+
+None of these four is settled. They are recommendations, not adjudications.
+
 ## 3. Expected workload
 
 Fifteen scored cases across three strata: roughly 12–16 material root causes,