# Three records, one word: completed · Tara · practise, chapter 2 · verification receipt · 11 September 2026

Scope: the second chapter of the _Tara · practise_ track. Twenty-one inherited
slides became twenty-seven: a cover, four section dividers, twenty-one content
slides and a close, in sections of five, six, five and five. The
[brief](../../authoring/tara-ritual-sessions-chapter.md) records what the
chapter is for and why each treatment was chosen; the track brief is
[tara-practise-track.md](../../authoring/tara-practise-track.md);
[state.json](state.json) holds the exact counts, tool results and findings.

## What was checked

- **Build freshness.** `build-eve-oshun.py --check` passes on the delivered
  tree: 1,483 slides and 663 sources. The chapter is 27 slides running 40.1
  minutes, with one table in it.
- **Every treatment was re-decided.** The plan listed three tables and seven
  card grids. One table survived, because it is the contract's own
  `SESSION_STATE_TRANSITIONS` map and a reader needs its rows side by side — and
  it grew from the six rows the inherited slide grouped them into to the eight
  the source declares, since `started` and `resumed` are separate keys and so
  are the two terminal states. The other two became card grids. One card grid
  became a compare panel, because its rows already asked one question twice of
  `completed` against `partially-completed`. The recovery map became a state
  machine, drawing the eight contract states as four boxes, each naming what it
  covers in its `states` field.
- **Bindings.** `assignments/v1-tara-sessions.json` binds nine of these slides
  across four handbook units. All twelve evidence entries carry fresh
  fingerprints, and each explanation is re-set to the rewritten takeaway of the
  slide it points at. `check-center-coverage.py --check` reports 347 reviewed
  and 0 stale over the whole library.
- **Browser rendering.** `eve-edition-review.mjs --guide tara-ritual-sessions`
  sweeps all 27 slides at 1440×900, 1280×720 with audio chrome, 390×844 and in
  print — 108 measurements, no spill, no document overflow, no page errors, no
  findings.
- **Narration.** Twenty-seven tracks rendered on this host's CPU and verified
  with `--check --ids`; the library is complete at 1,974.0 minutes with nothing
  pending.
- **PDF.** 27 pages, exact authored visual text on every one, in order, read
  back from the downloadable `decks/tara-ritual-sessions.pdf`, and
  `check-pdf.py` passes including its table-header size floor on the transition
  table.
- **Suites and lint.** Python 473 passed, 0 failed. Node 489 of 489.
  `ruff check` and Prettier clean on every touched file.

## The capture, and the path it was first through

This chapter has no surface of its own — it is contracts and library functions —
so its one capture rides on a case study. The composition contract has allowed a
capture beside a case for as long as the Eve edition has existed, with a
validator, a renderer, a stylesheet rule and three unit tests; no delivered
slide had ever used it. `tara-session-percent-case` is the first.

It carries the practise track's own pinned player crop, at the pin
`1ceae52c7896a59ac74a4429d410f3218f0b91c6`, because the slide is about a
percentage and the only percentage a member currently sees is a third one again:
the player's elapsed seconds over its planned duration, reading zero on a fresh
mount. Neither the ritual meter nor the completion decision appears on any
shipped screen.

Being first through the path found two things, and both are fixed here rather
than noted:

- `.case-capture img` set `width: 100%` with a capped `max-height`, so the
  bordered frame drew wider than the picture inside it and left dead space to
  its right. The width follows the height now, capped at the container, and the
  border sits on the image. The rule's own comment had already described the
  intended behaviour; the width defeated it.
- A composition guard asserted that no delivered case study renders a capture.
  That is exactly the regression it was written to catch — a restyle quietly
  attaching pictures to existing cases — so it is not deleted. It names this one
  slide as the deliberate exception, and now checks both directions: a slide
  that renders a capture must declare `media`, and a slide that does not must
  not.

## What was fixed

- The case study carrying the capture ran 26px past the page at 1440×900, where
  the capture's `26vh` cap is tallest. Its subtitle, case detail, caption and
  two node details are shorter; it passes at every viewport now, and at 1280×720
  it never failed.

## What was not checked

- No persistence, delivery or notification. Every function in this chapter
  returns data: the contract validates a supplied record, the evaluator returns
  a decision, the skip and recovery helpers return new objects, and the event
  and projection builders return more objects. None of them saves or sends, and
  the chapter says so on each slide rather than once at the end.
- No authorisation. An actor field is a declared string, a memory scope is a
  declared purpose, a consent reference is a nullable identifier nothing reads,
  and a resume token is routing context built from a member identifier, a plan
  slug, a week and a day.
- No deployed consumer for the four completion projections, and no query-time
  subject filtering. The inspected lifecycle integration test composes assembly,
  scheduling, completion and projection over fixtures.
- No clinical or personal claim. The three-step ritual, its durations and its
  progress are illustrative evaluator input.
