# What to practise, and who guides it · Tara · find a practice, chapter 5 · verification receipt · 11 September 2026

Scope: the fifth and last chapter of the _Tara · find a practice_ track, and the
first delivery in this programme that retires a guide. Forty-seven inherited
slides across two guides became forty-two: a cover, five section dividers,
thirty-five content slides and a close, in sections of seven. `tara-selection`
kept the chapter id and the twenty-two slides of `tara-teachers` folded into it,
as the track brief said they would. The
[brief](../../authoring/tara-moods-teachers-chapter.md) records what the chapter
is for and why each treatment was chosen; the track brief is
[tara-find-a-practice-track.md](../../authoring/tara-find-a-practice-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: 1,470 slides and 663 sources. The chapter is 42 slides running 60.6
  minutes, with three tables in it.
- **Every treatment was re-decided.** The plan listed fourteen slides as
  `evidence-comparison → card grid`, and they were fourteen instalments of three
  catalogs nobody ever saw whole: six mood slides, three theme slides and five
  modality slides. They are three exact registries now — twelve moods, fifteen
  themes, fourteen modalities — and every value is still on a slide. Three other
  slides left the plan: a compare panel whose rows already asked one question
  twice of two nested records, a stat panel that is three counts of one word,
  and a third capture, because a Play icon on a row with no handler is a visual
  claim.
- **Fifteen slide ids were merged away** and registered in `MERGED_AWAY_SLIDES`
  with the slide that absorbed each one. Fourteen went into the registries; the
  fifteenth is `tara-teacher-opening`, folded into `tara-select-close`, because
  two guides opened with a case each and a chapter opens once.
- **Bindings.** Both `v1-tara-selection-safety.json` and `v1-tara-teachers.json`
  were re-pointed at the merged guide with fresh fingerprints. The mood, theme
  and modality units now cite a registry where they used to cite instalments,
  with new explanations; the Teachers unit keeps its seven slides and changes
  only its guide id. `check-center-coverage.py --check` reports 347 reviewed and
  0 stale over the whole library.
- **Browser rendering.** `eve-edition-review.mjs --guide tara-selection` sweeps
  all 42 slides at 1440×900, 1280×720 with audio chrome, 390×844 and in print —
  168 measurements, no spill, no document overflow, no page errors, no findings.
- **Narration.** Forty-two tracks rendered on this host's CPU and verified with
  `--check --ids`; the library is complete at 1,955.2 minutes with nothing
  pending.
- **PDF.** 42 pages, exact authored visual text on every one, in order, read
  back from the downloadable `decks/tara-selection.pdf`, and `check-pdf.py`
  passes including its table-header size floor on all three registries.
- **Suites and lint.** Python 473 passed, 0 failed. Node 489 of 489.
  `ruff check` and Prettier clean. No JavaScript changed.

## The registries, and what it cost to fit them

The library's longest comparison table before this chapter was six rows. At the
comparison row height — 63px at 1440×900 — a twelve-row table is 436px of a
530px visual band and a fifteen-row one does not fit at all, and at 1280×720
with audio chrome the band is 411px and the controls get pushed off the screen.

So a table longer than a comparison now declares `density: "registry"` and takes
a compact row: 3px of vertical padding instead of 17, and
`clamp(13px, 1.02vw, 15px)` instead of `clamp(16px, 1.4vw, 21px)`. The validator
requires the declaration above six rows and forbids it at or below six, so the
choice is in the data rather than in a threshold nobody can see. Nothing already
in the library is affected — no delivered table has more than six rows.

The header does **not** take the compact size. The exported PDF is checked for a
12px effective table-header size at a 960px page fit, which is exactly the
library's 16px header, and the first export failed on it because the compact
rule outranks the bare `thead th` selector and the header had silently inherited
13px. It is restated explicitly now, with the reason.

## The captures, and where they came from

Three, all from `assets/tara-teachers/`, all pinned at
`1ceae52c7896a59ac74a4429d410f3218f0b91c6` and rendered from the actual
components over their shipped seed data. Two were already cited by the inherited
`product-view` slides; the third had been captured in September and never used.

- **The directory** under the Anxiety filter, four pins: a seeded rating and
  session total, the featured flag that wins outright, the card's three-label
  limit, and the Active threshold on the neighbouring card. The two badges
  differ for a reason the picture cannot show, which is what the legend is for.
- **The lineage disclosure**, three pins: the badge that says Not yet verified,
  the future path it names, and the absent citation trail it explains. This is
  the most accurate element on the shipped profile and the chapter says so.
- **The session list**, three pins: a Play triangle on an `article` with no
  handler and no href, a rating that belongs to the session rather than the
  teacher, and the three rows the seed join actually returned.

Every pin was placed by scanning the PNG for the nearest centre whose
twenty-eight-pixel radius is clear of ink, not by estimating one from a
rendering.

## What was fixed

- A stat panel's `exampleLabel` was authored and never painted — the renderer
  builds its own label from `kind` and the shared source path. The sweep caught
  it at all three viewports. The field is removed; the panel says
  `Configured values` and prints each number's own source path, because these
  three counts come from two different files.
- The `Finding a teacher in the directory` divider's subtitle ran to two lines
  at 1440×900 and pushed the narration chrome 20px off screen. It is shorter.
- The mood registry's handoff column read `required · start suppressed`, which
  widened the column enough to wrap four of the twelve rows and put the chapter
  5px over at 1280×720. The column says `required`; the suppression is in the
  notes and on the base-state card grid, where it already was.
- The modality registry's notes said ten of the fourteen carry a
  contraindication note. It is twelve — every entry except Silent and Guided.

## What was not checked

- No human reader session and no listening review at normal speed.
- No live teacher service, no Sophia verification read, no collected customer
  reviews and no booking. The twelve teachers, their ratings, their session
  totals and their certifications are development seed data; the reviews are a
  deterministic hash of the teacher id and the review index; the weekday grid is
  a hash of the weekday name alone and is identical on every profile.
- No clinical claim. Distress levels, contraindication notes and crisis handoffs
  are product mechanisms with schemas behind them. The three base-high slates
  are shown because leaving them out would misrepresent the catalog, and every
  row that shows one shows the required handoff beside it.
- No application caller. Nothing in the three catalogs composes the three reads,
  and the chapter says so on the slide rather than in a footnote. The Lilith
  handoff guide follows permission separately.
