# W05–06 targeted improvement · actual QA and freeze record

**Date:** 2026-09-30. **Status:** independently desk-reviewed teaching draft; final manifest/check commands below passed. This is an author-agent internal desk/scripted review, followed by the separate publishing-agent review in [RUN-THROUGH.md](RUN-THROUGH.md). No actual teacher, parent, child or funder took part; no classroom pilot, human audio listening, intended-device/volume approval, accessibility certification or year-coverage certification is claimed. All six candidate recordings remain **NOT CLEARED** and excluded from playback until local human gates.

## Exact scoped change

Only this W05–06 folder is authored. Ordinary Days 21–24/27–29 now offer a short actual familiar locally reviewed quiet phrase before score comparison, own purpose/music alone, private one-part making and meaningful keep/change/stop/rehearsal. Direct-adult sounding and silent plans are complete for their stated evidence; actual child performance remains separate. A tap cannot realise LONG. Unsuitable/unavailable sustained sources lead to a separate short/rest actual trial plus silent held plan, not a false hold claim.

README starts with a practical tiny opening/ending invitation. Materials/routes/prompts/home offers are clearer; all 30participation routes,20 worked swaps and20 free optional bridges remain. No new formal check, mandatory recording, purchase, specific voice/body/source/control/rate/speed, simultaneous role or public performance. No interactive extension added: the child-owned real phrase and one existing print kit serve the immediate music task without extra code exercises.

Little Window Rhythm supplies an original named SubjectNest computer-created saved graphic model/source, accurately dated 2026-09-30 for its naming/source/new synthesized versions, with the prior A pattern acknowledged. Its documented author purpose is distinct from achieved effect. Actual source encounter and own child response offer only conditional narrow FE01 evidence; varied appropriate people/community works and local authority remain needed. Three new original pulse-only/rhythm-only/together candidates distinguish the two parts. Together has seven individual part-events at four distinct onset times, never a compulsory duet.

## Preserved fresh material

[Pinned baseline](PRESERVED-BASELINE.json) and verifier fixed constants confirm **byte-for-byte unchanged**: separate teacher key; K and L SVG/PDF; original H/I/J WAVs. Practice B/C/D/E SVG/PDF, worked swaps, authentic-source slot and font notices have no diff. Public STUDENT-CHECKS retains all original data/wording with exactly one added timing paragraph: unfinished = **not yet observed**, another short sitting, no completion-speed score. Both fresh lessons add the same timing boundary. No new practice/media reproduces K's1/2/4 answer or L'sX1/LONG2–3/REST4 score. Check-day home offers use unrelated **After Check** practice rows.

## Actual source checks

[Receipts](SOURCE-RECEIPTS.json): fresh direct ACARA workbook **HTTP 200, 2,049,917 bytes**, SHA-256 `db446882d2c00cf7c085a03e250e2442fda6c44011fc114680c46c1dc7a822c3`, equal to pinned canonical workbook. Verifier reads actual Learning areas E/J rows 21218/21227/21237/21246 and imported Foundation Music descriptions, plus Foundation Arts achievement-standard row 12. It does not substitute Year 1–2 Music. Fresh official indexed QCAA alignment one-page text and May 2026 example four-page text were read; both direct downloads **403**, no PDF downloaded. Current ACARA Arts overview text was read. No new successful Music-examples/terms-page fetch is invented. [Crosswalk](CURRICULUM-CROSSWALK.md) and [rights](SOURCE-AND-RIGHTS.md) retain exact codes, scope, official terms and First Nations local authority limits.

## Mechanical and substantive checks actually run

| Check | Actual result / scope |
| --- | --- |
| Ten daily scripts | All 10 have exactly3+4+5+7+4+2=25 minutes; seven improved ordinary blocks have practical trial before score discussion and child keep/change/stop. Short sittings remain permitted. |
| Route/example/check integrity |30distinct choice routes;20 worked swaps;20 free optional home offers;two held fresh checks; separate unchanged key; existing planning answer consistency retained. |
| Printed music data | Actual SVG positions verified independently for A/B/C rhythm event/rest places and D/E held-band/short/rest positions, not merely counts in expected arrays. Fresh asset hashes fixed. |
| Eight actual PDFs | Each one-page A4, searchable, embedded DejaVu; all actual word bounds inside usable page, no word overlap. Full pages rendered with Poppler at 100 dpi and visually inspected: A/F, B/C/D/E, K/L. No observed cropping/clipping; revised source caption and own-work fields readable. |
| Exact alternatives | Every SVG printed word appears verbatim in matching linear/tactile alternative; full row order, rest and one-band meaning supplied, with sound/evidence limits. |
| Six actual candidate signals | Original H/I/J retained and checked; all three new signals reconstructed sample-by-sample from independently fixed score/recipe. Full event timing, component identity, silence, valid mono 16-bit/22,050 Hz/46,305 frames/2.10s format and no clipping verified. Digital peaks H/I/J 1802, new pulse 393/rhythm 409/together 801; none predicts acoustic comfort. |
| Ten negative probes | Rejected same-count fresh-answer practice leak; wrong actual rest; changed fresh key; shifted pulse timing; missing rhythm despite same four onset places; incomplete silence transcript; misplaced held band; false human attribution; silent direction counted adult performance; two-page aid. All in temporary copies, original unchanged. |
| Byte reproduction | All 24generated files reproduce exactly in a temporary directory:16 SVG/PDF,6 WAV,1 exact-alternative file,1 font notice. Owned pack unchanged by reproduction. |
| Source/code and formatting | All four exact official rows against raw workbook/import; source receipts/fetch limits; rights/local authority and evidence boundaries; Python checks with no cache; scoped diff whitespace check. |
| Default verifier/read-only | Full verifier and default hash audit pass; disposable __pycache__/.ruff_cache/.pytest_cache excluded. Manifest covers48 resource/source/QA hashes, excluding itself. Read-only/probe/reproduction runs preserve the pack bytes. |

## Four-role desk critique

[Full run-through](RUN-THROUGH.md) records concrete teacher, learner, parent and assessor/funder viewpoints. Teacher: start with one page/source and one phrase; actual makers/holds recorded privately. Learner: choose purpose/preferred ending, one part/private route and stop, without formal audit language. Parent: one minute/free invitation or none, with no return report. Assessor/funder: canonical/source/media/key integrity supports honest partial evidence, not a classroom benefit or whole-year completion claim. These are internal analytical roles, not observed participants.

## Portable review commands

From this pack directory:

```sh
python3 verify_pack.py
python3 verify_negative_probes.py
python3 verify_reproduction.py
ruff check --no-cache generate_assets.py verify_pack.py verify_negative_probes.py verify_reproduction.py
```

Intentional regeneration uses `python3 generate_assets.py --write`, followed by actual layout/data inspection and `python3 verify_pack.py --write-manifest`. Reproduction requires Python 3, rsvg-convert, DejaVu fonts and Poppler tools available; negative two-page probe also uses pdfunite. These commands contain no private workstation paths.

## Gates still open

Human hearing/device/room/comfortable volume/stop clearance for each candidate; local review of actual live sources/access; actual child making, aural response, pulse/rhythm, work encounter/response and willing audience observations; varied appropriate repertoire/cultural authority; classroom trials, educator/parent feedback and independent accessibility review. An invitation, saved plan, computed signal or generator completion supplies none of those human observations. No staging, commit, push, shared-map edit or other pack edit by this improvement leaf.