# Australian taxonomy cleanup — 23 October 2026

Applied through the authoritative taxonomy builder and staging database on 23 October 2026. This local export matches the independently verified staging checkpoint. Earlier Australian authoring reports describe historical snapshots.

## Reason for the cleanup

The retrospective reviewed all 314 Australian additions and 14 later edited IDs (five overlap the additions, giving 323 distinct records). It identified repeated skills authored for different grades, media, or contexts, plus bundled mastery requirements and some genuine additions.

The initial review classified 160 additions as distinct, 45 as merge candidates, 39 as extensions before possible retirement, six as localization before possible retirement, 47 as splits/redistributions, and 17 as uncertain. Those are original review categories, not final deletion counts. Subsequent discussion and delegated decisions gave all 17 uncertain cases a treatment direction.

## Coverage interpretation

All 593 selected Australian expectation codes and 1,848 original topic–code associations were mechanically verified at the audit baseline. That establishes recorded associations, not coverage of every clause. The retrospective independently checked 21 expectations clause by clause; nine had alignment or proposed-transfer issues. The other 572 previous Full claims were not independently recertified.

Earlier reports' Full labels are historical authoring judgments. The cleanup checks affected clauses, records whole/component/prerequisite distinctions, preserves genuine knowledge and operations, and transfers requirements before retiring duplicate IDs. Current mapping counts and remaining limitations are recorded below and in the linked verification records.

## Mastery and learner records

Each retained topic has one mastery claim; this cleanup does not introduce separately tracked difficulty levels. Distinct number-system demands, procedures, and bodies of knowledge can remain separate. Grade, country, or software medium alone does not establish a new skill.

Learner evidence is preserved separately from duplicate age-derived priors. Priors are not added or averaged during consolidation. The user explicitly approved neutralizing 52 inherited estimates on two expanded number topics pending reassessment; the actual learner observation and all assessment history remain intact. Private learner backups are not included in this repository.

## Applied outcome

| Original additions | Final count |
|---|---:|
| Retired after clause redistribution | 106 |
| Retained unchanged | 130 |
| Retained with narrowed claims | 49 |
| Retained with other edits | 29 |
| Total original cohort | 314 |

Seven focused additions are recorded separately: Taking action to include others (`mt_1MqcYKS1Es`), Cutting with a saw safely (`mt_8FUXG_39-C`), Reading nutrition labels (`mt_CCJpSSS3l8`), Joining materials with adhesives (`mt_EiI2PPWPHJ`), Sewing by hand (`mt_FfMBoU1jt7`), Cutting food with bridge and claw grips (`mt_V6SzjffDoE`), Drilling holes safely (`mt_uUMLgRTrhn`). The result is 99 fewer topics than the audited 1,932-topic staging baseline: **1,833 topics and 3,846 prerequisites**.

The [314-row execution ledger](acara-v9-execution-ledger-2026-10-23.json), [14-edit ledger](acara-v9-existing-edit-ledger-2026-10-23.json), [106 retired-ID mappings](acara-v9-retired-topic-mappings-2026-10-23.json), [nine source-issue closures](acara-v9-source-issue-closures-2026-10-23.json) and [593-code status matrix](acara-v9-source-status-2026-10-23.json) preserve the review trail. A component association is not a whole-standard certification. Optional medium/tool examples are not invented curriculum obligations.

## Verification and limits

Independent readback compared every persisted topic field, country/age alignment, prerequisite and resolvable standard association with the authoritative graph. Normalized reachability centrality matches; no topic embedding is missing. The graph is acyclic and introduces no new hard-age or missing-country prerequisite conflicts. Existing baseline conflicts remain outside this cleanup; this is not a global certification of prerequisite quality. The 41 historical unresolved associations are unchanged and excluded from resolvable public links.

All 17 source catalogues and 7,743 expectation payloads are unchanged. Ontario retains 172 associations across the same 55 codes after documented composition/data-collection transfers and removal of two weak analysis/prediction associations. Nine Ontario documents add 3,862 source expectations to this local export as codes and official links; catalogue ingestion does not mean all Ontario expectations are mapped. The focused authoring pass was Grade 1 Mathematics.

The private preservation comparison confirms actual learner observations/evidence, favourites, interests, tinker references and assessment history are unchanged. Retired age-only priors were not combined. The only existing learner-estimate changes are the 52 explicitly approved resets. Three teaching-research records have the documented nutrition or number-assessment corrections; historical research provenance and the two archived records remain preserved. No learner payload is included here.

Dataset schemas, referential integrity, research coverage/source links and archive checksums pass, along with 14 dataset regression tests and nine exporter tests. The exporter preserves the existing historical exclusions and assessment-reference redirect; it fails closed if an archived source record has changed. This is a local export update, with no production promotion or public push. The 572 historical Full claims remain unrecertified.

PR review corrected the across-100 comparison wording and stale fabrication replacement entries. The corrected evidence and teaching guidance use the highest relevant place first; sawing, drilling, adhesive joining and sewing have separate recorded destinations. Staging preservation checks were repeated without changing learner state.