# Eve field guide · completion evidence · 8 September 2026

The remaining Eve design pass is delivered across **25 guides / 204 slides**.
The [sequence review](../../authoring/eve-design-review.md) explains the order of
all 25 guides. The [slide ledger](../../authoring/eve-quality-ledger.json) binds
all 204 retain/change decisions to their expanded manuscript hashes.

## What reached the slides

- Thirty-three concrete or corrected product explanations, four enlarged
  architecture graphs, and the remaining category-header correction are recorded
  in [the intervention list](../../authoring/eve-design-changes.json).
- One Enrollment investigation connects feature graphs, direct coverage, work
  capture, decisions, delivery and the closing exercise. Recovery teaches all
  five legal exit/recovery groups and their exact state transitions.
- Eleven assets render the actual pinned Explorer, Work and Audit React UI with
  declared fixtures. The capture harness, source hashes, fixture, PNG hashes and
  executed readiness-evaluator result are in
  [capture provenance](../../assets/eve-product/provenance.json). These are
  controlled examples, not observations of a deployed backend.
- Worked records explain ADR replacement direction, leases and fencing,
  verification outcomes, changed-file manifests, incident investigation, Studio
  dossier limits, readiness gates, scheduling and supported member operations.
- All Eve slides use the quieter field-guide typography and remove the repeated
  masthead. Independent columns no longer imply numbered steps. Category-only
  headers do not say “Question”; genuine question columns retain that label.
  Highlighting is explicitly authored for a meaningful claim, never selected by
  row position.
- Keyboard panning inside a product image no longer advances the deck. On paper,
  the Work slide enlarges the in-review task while native annotations retain all
  ten states; on screen the complete board remains pannable.

## Review and automated evidence

| Evidence | Actual scope and result |
| --- | --- |
| `published-matrix/report.json` | 1,434 checks: all 204 slides in four standalone and three embedded views, plus six enlarged-text checks; no errors or fit findings. Axe runs on all three embedded views. Last-track playback for all 25 guides. Three negative controls rejected. |
| `print-detail-browser/report.json` | 167 checks after the final picture/print change: all 23 slides in the three product-image guides across the same view matrix, plus enlarged text. No errors or fit findings; three negative controls rejected. |
| `capability-browser/report.json` | 282 source/text/reflow checks over all 94 capability slides, plus four negative controls. |
| `architecture-browser/report.json` | 148 source/text/reflow checks over the seven detailed architecture guides, plus four negative controls. |
| `negative-controls/report.json` | Six known heading, positional-emphasis, edge, refusal and quantitative defects are rejected. |
| `zoom/report.json` | Actual Chrome tab zoom at 200% and 400% for five difficult subjects: ten passing checks. |
| `bank-layout.json` | All 578 unique slides in desktop, compact and mobile layouts; all 578 printed pages; complete table-field and header floors. This `--layout-only --draft` run does not claim per-slide axe or every-track playback. |
| `center-smoke.json` | All 60 guides navigate; search/history, mobile focus, narration controls and navigation performance pass. |
| `final-exports/report.json` | Final 25 PDFs / 204 pages, exported after the Work print enlargement; print-image selection and load completion checked. |
| `capability-pdfs.json`, `architecture-pdfs.json` | Exact ordered PDF text for 94 capability pages and 50 detailed-architecture page instances. Extraction alone does not establish optical legibility. |
| `logs/units-final.log`, `logs/print-unit.log` | 125 focused Python tests passed, followed by 39 affected composition, provenance and PDF-meaning tests after the print correction. These counts overlap. |
| `logs/ruff-*.log`, `logs/eslint-final.log` | Relevant Python and JavaScript lint passed. |
| `logs/builder-final.log`, `logs/assignments-final.log`, `teaching.json` | Builder freshness, source/narration integrity, reviewed assignments and capability crosswalk passed. The library contains 60 guides, 578 unique slides and 348 cited paths. |

The final optical review inspected all 25 browser sequences and all 25 final PDF
sequences, with native-size inspection of changed and dense subjects. `sequences/`
and `pdf-sequences/` retain those contact sheets; `pdf-pages/` retains selected
native PDF renders. `retained-captures.json` identifies the selected native browser
captures actually retained here. The reports also name captures from the complete
runs that were not copied into this receipt. Selected rejected Work views are
kept under `rejected/` when available.

This is model editorial/optical review, not human comprehension evidence. Neutral
reader tasks and scoring remain in [reader review](../../authoring/reader-review.md).
No new willing-reader sessions were performed or invented. The separate
whole-library teaching/design program and live Eve editing service remain open.

## Final-change continuity

The complete Eve matrix preceded the final print improvement. That improvement
adds two print-caption visibility rules and a `picture` wrapper in the three
product-image guides. `screen-reviewed.css` and `screen-reviewed-work.json` retain
the exact earlier inputs. `verify-receipt.py` proves that the other 22 guide HTML
files reproduce their tested hashes when the appended CSS is removed, and that
Work's only manuscript differences are its print image and image alternative.
The current 23-slide product-guide run covers all three renderer changes. All 25
PDFs and the zoom run were regenerated/rechecked after that correction. This
bounded proof avoids repeating the unchanged full-library browser run.

The earlier capability/detail text checks preceded JSON formatting. The retained
`reviewed-json/` files reproduce their exact tested hashes; the format proofs
check parsed equality against delivery, allowing only the separately tested Work
print addition. Historical report hashes are preserved rather than relabeled.

`state.json` binds current authoring/rendering/test inputs, all 60 guide HTMLs,
all 25 Eve PDFs, all 578 current narration tracks and retained evidence. Thirty-two
assignment links were reviewed after the style changes; `assignment-review.json`
records their substantive comparisons. The parallel telemetry update from
`899e3856cc` and its regenerated recording were preserved and verified.

## Reproduce

From the repository root:

```sh
python3 docs/presentations/presentation-center/verification/eve-completion-2026-09-08/verify-receipt.py
python3 tools/presentations/build-eve-oshun.py --check
python3 tools/presentations/check-center-coverage.py --check
```

The runnable browser and capture harnesses are under `tools/presentations/tests/`:
`eve-completion.mjs`, `eve-product-capture.mjs`, `eve-depth.mjs`, `eve-zoom.mjs`,
`eve-design.mjs`, `eve-design-export.mjs`, `eve-oshun-deck.mjs`, and
`presentation-center.mjs`. Their output-directory and scope arguments are declared
in each harness. Run expensive checks serially through `run-supervised.py` after
checking resources. The completed logs above are the evidence for this delivery;
a future content/rendering change requires fresh affected evidence.
