# Tara service discovery, saved items and launch · Tara · find a practice, chapter 4 · verification receipt · 11 September 2026

Scope: the fourth chapter of the _Tara · find a practice_ track. Twenty-four
inherited slides became thirty-one: a cover, four section dividers,
twenty-five content slides and a close, in sections of six, five, eight and
seven. The [brief](../../authoring/tara-discovery-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. The chapter is 31 slides running 41.9 minutes, with one table in it.
- **Every treatment was re-decided.** The plan defaulted twelve slides to card
  grids again; four of them are card grids here. Three became compare panels
  whose rows already asked one question twice — what a role declares against
  what it receives, what an object says against what its action does, and how
  two entry points treat the same duplicate. One became the chapter's only
  table, because three scoring tables of three numbers each are the rule itself.
  One became a layer stack, because build, filter and slice is an order with a
  rule at each boundary. One became a timeline, because the page walk is a
  bounded sequence along offsets where each point decides something. And four
  gated flows became decision trees that each kept two outcomes, because the
  source has no third branch in any of them.
- **Bindings.** No assignment file and no teaching guard binds a slide in this
  chapter; every `assignments/*.json` evidence entry was checked by id and the
  node suite is green on the rewritten manuscript.
- **Browser rendering.** `eve-edition-review.mjs --guide tara-service-discovery`
  sweeps all 31 slides at 1440×900, 1280×720 with audio chrome, 390×844 and in
  print — 124 measurements, no spill, no document overflow, no page errors, no
  findings.
- **Narration.** Thirty-one tracks rendered on this host's CPU in batches of six
  and verified; the library is complete at 1934.6 minutes with nothing pending.
- **PDF.** 31 pages, exact authored visual text on every one, in order, read
  back from the downloadable `decks/tara-service-discovery.pdf`, and
  `check-pdf.py` passes including its quantitative row check on the moment
  table.
- **Suites and lint.** Python 473 passed, 0 failed. Node 489 of 489. `ruff
  check` and Prettier clean.

## The capture, and where it came from

The chapter has no surface of its own. Its last section spends five slides
describing a destination — an action, a target path, a query and a parsed
intent — so the capture is that destination: the practise track's own pinned
capture of the shipped sitting player, cited here rather than re-rendered.

Three pins. The five-minute sitting the path names. The controls that nothing in
this chapter pressed. And the ambient sound request the player reports as
refused rather than playing silence and claiming success — which is the concrete
reason a resolved path is not a completed journey: the path was fine, and the
thing at the end of it was partly unavailable.

## What was fixed

- The layer stack was authored with a `contents` field on each band. The browser
  sweep expects that field to be painted — it is in its node-text list — and the
  renderer paints only the title and the detail, so three authored strings were
  invisible at every viewport. The contents are folded into the details now. No
  delivered layer stack uses the field; the sweep's list is ahead of the
  renderer, and this is the first slide to find out.
- The moment table's row label wrapped to two lines, and the PDF coverage check
  reads quantitative rows through a layout extraction where a wrapped label
  breaks the row apart. The label is shorter; the note already names which
  moments it covers.
- The capture's third callout landed on the notice it was pointing at, at the
  scale the figure is drawn. It was repositioned by scanning the image for the
  nearest centre whose thirty-pixel radius is clear of ink, rather than by
  estimating one from a rendering.

## What was not checked

- No human reader session and no listening review at normal speed.
- No downstream transport, database, token verification or playback. The route
  exercises run real Fastify routing with a declared context and a no-op abuse
  prehandler, so they verify the route guard and not the middlewares they stand
  in for.
- No application consumer of the favourites page walk. A pinned symbol search
  found the helper, its tests and a documentation reference, and no call site in
  the searched scope; the chapter says so on the slide rather than in a
  footnote.
