# Four-role author desk critique · 30 September 2026

**Evidence status:** these roles were simulated by the author while checking the actual files and rendered pages. No independent teacher, student, parent/carer, cultural authority or funder participated. No child or actual location data was collected. This is a draft review, not a classroom or accessibility trial.

**Teacher role:** I built the source relationship as two paper-model instructions: A earlier basket by rug; B later reported basket by table. C matched A's drawn location but stayed outside known-time order. D gave Tara's later reason without filling Remy's cause. E was a picture proposal, while the learner's own revised map could be a real child artifact without implying a real basket move. I read all ten timed scripts and checked that their products progress through bases, time windows, undated pocket, named message, planned map, revision and gallery. Fresh check source numbers are neutral before first response. The actual-place option contains five complete replacement sessions, preflight, question stems, record fields and element-specific K03 decisions. Its current/previous fields remain separate, so a teacher with no previous source does not invent a before view. Local lesson pacing, source language and curriculum decisions remain educator review gates.

**Student role:** I followed a print/point route, an exact oral-description route and adult-directed broad-piece placement in a desk simulation. The same source goals survived the response change. I kept the earlier model, built a later reported model and put the undated picture in a separate pocket. I could attribute Tara's meaning and keep a proposed map separate from physical action. With the new Moss/Harbour sources, I could distinguish unknown copy time, named reason, keeper cause and actual child-model making. These are author answers, not evidence of any child's comprehension, interest or independence. Raised outlines were specified and walked through verbally; none was physically built or learner-tested.

**Parent/carer role:** I checked that each bridge uses free, optional invented paper play with an in-school equivalent and no home inspection, real map, address, family interview, device, recording, answer return or adult availability. A pass has no assessment cost. The actual-place option is school-approved work, not a request for private home evidence. A home language may carry reasoning; any supplied translated version would require fluent human/local review. No parent/carer reviewed the activities.

**Potential funder role:** I checked for ten timed practical lessons, thirty switchable routes, twenty worked everyday-domain variations, ten bridges, two fresh checks with separate key, nine original print pairs/alternatives, five runnable actual-place replacements, exact curriculum pins, rights and complete hashes. These support authored-draft completeness and reproducibility. They do not establish teacher time saved, demand, learning gain, classroom participation, completed K03 knowledge, actual care benefit, full-year coverage or commercial suitability.

**What was actually inspected:** all nine PDF pages were rendered locally and viewed together. Shape keys, earlier/undated diagrams, reported-later positions, source text, blank actual-place prompts, panels and footer remained visible without clipping. The Harbour sheet's last two panels were resized before acceptance when generation detected insufficient text-panel space. Measured text bounds, embedded/searchable PDF text, exact SVG/PDF/alternative correspondence and A4 geometry are checked in the verifier. The current official workbook HTTP 200 hash matched its pin, and the current official QCAA one-page Prep alignment matched all nine descriptions. Actual place/access/safety/consent/care evidence has not been collected.

**Residual critique:** the undated-source distinction is demanding for Foundation. Teachers should teach one source/time relationship at a time and ask short prompt parts separately. Physical model making may consume the seven-minute inquiry; pre-make broad bases/pockets and allow complete adult-directed construction. A source report can be mistaken; the task asks what it reports, not whether real physical conditions are proven. One small actual-place inquiry may provide partial K03 evidence by element, while missing belonging, personal meaning or care fields stay open. PDFs are untagged; screen-reader order, print legibility, tactile construction and individual AAC access remain human review gates. No “10/10” quality or child outcome claim is made.

**Reproduction:** from this pack, `python print/generate_print.py --regenerate` deliberately rewrites only nine named SVG/PDF pairs and alternatives. `python verify_pack.py --manifest-json > manifest.json` deliberately saves a proposed file receipt; normal `python verify_pack.py` is offline and read-only. Python 3, Pillow/DejaVu, librsvg, pdfinfo, pdftotext and pdffonts are required. The fixed PDF build clock is reproduction metadata, not an actual event/source date.

## Final verification receipt · 30 September 2026

- **Offline audit:** PASS — nine exact direct workbook rows/QCAA pins; ten 25-minute scripts plus five actual-place replacements; conditional K03 with no outcomes observed; 30 routes; 20 worked swaps; 10 bridges; two separate fresh checks; nine A4 pairs/exact alternatives/text bounds/keyed geometry; 59 local links; 37 SHA-256 files. Every required file except the manifest itself is hashed.
- **Reproduction:** deliberate second regeneration gave byte-identical nine SVGs, nine PDFs and the full alternatives file: 19/19 artifact hashes matched. The fixed PDF metadata clock is for reproduction only.
- **Negative checks:** 12/12 in-memory mutations were rejected: workbook digest, K01 hold, blanket K03 hold, invented K03 outcome, daily timing, actual-route timing, missing no-actual-evidence statement, fresh-speaker leakage, learner/key leakage, exact alternative drift, stale receipt and shifted keyed map geometry. Each original was restored in memory; no bad fixture was written into the pack. Normal audits before/after passed and every file's bytes, size and modification time were preserved by these exercises.
- **Print/source inspection:** all nine rendered pages were visually inspected; no clipping was found after the Harbour panel-spacing correction. Exact SVG/PDF text/alternatives, text bounds, source quotation correspondence, source-backed icon placement, A4 page size and embedded searchable fonts passed. The official workbook re-fetch matched its SHA and the official current QCAA Prep alignment matched all nine descriptions.
- **Code/runtime:** both original Python tools pass formatting and lint checks. Python 3.12.3; librsvg 2.58.0; pdftotext/pdffonts 24.02.0; pdfinfo 26.05.0; DejaVu Sans fonts. Normal verifier operation uses no network, writes no pack file and collects no learner data.

**Remaining human gates:** educator language/pacing and school adoption/timetable; actual-place approval, privacy/consent/care procedure; individual observation/communication/tactile/print access; cultural authority/ICIP where applicable; evaluated classroom use. No actual K03 evidence or care action was collected. K03 remains a runnable conditional opportunity, while K01/K02/K04 require separate authentic teaching/evidence. No classroom, access, cultural or commercial quality certification is claimed.