# Tara ritual assembly and tone review · authoring record · 9 September 2026

## Purpose and scope

This package continues the Tara member-experience route for product, design,
engineering, editorial and governance readers. Its observable outcome is to
assemble one ritual, explain every rule and failure edge involved, and decide
what evidence is still missing before a template can be called publishable.

The guide follows `tara-context-lineage`, `tara-ritual-sessions` and
`tara-scheduling`. It reviews the complete handbook units **Rituals, steps,
triggers, and scheduling** and **Contemplative tone review**. Partial
completion, audio, companion, teacher, Living Scenes, Mentor Presence and
workbench units remain separate work.

## Baseline and source boundary

Primary documentation remains pinned at
`1ceae52c7896a59ac74a4429d410f3218f0b91c6`. The implementation, presentation and
narration preservation baseline is `dec34aacff0f40e368e5caf412d5df43df4a6a41`.
The complete 17-file source set—one handbook plus sixteen implementation and
test files—is byte-identical across the source pin, baseline and current
worktree.

The review covers the handbook units, canonical Tara contract schema, local
domain ritual assembly/model, Nisaba and Nyx relationship dependencies,
`@tara/features` ritual-template builder/validator, Lilith Tara policy binding,
deterministic tone evaluator and their owning tests. Twenty pinned cases execute
the rule catalog, assembly, failures, duration budgets, all three template
shapes, tone selection, scoring, publish gate and stale-review invalidation.

## Source review findings

Three related template records are not interchangeable:

- `TaraRitualTemplate` in the local domain package assembles five step kinds,
  cross-domain destinations, completion defaults and local taxonomy.
- `RitualTemplate` in `@tara/features` carries authored instructions, triggers,
  scheduling, rights, accessibility and a Lilith tone-review record.
- `RitualTemplateSchema` in `@oshun/contracts` adds canonical mood, theme,
  modality, lineage and duration bindings through Zod validation.

No converter or application consumer linking those planes appears in the bounded
direct-symbol search. The guide therefore teaches the records and verified pure
functions without inventing an end-to-end publishing pipeline.

Assembly chooses the first available anchor in each moment's ordered
`anchorKinds`; a second supplied anchor is omitted. Required passages can be
derived from the Nisaba relationship map. An explicit `null` refuses a required
passage. Reflection-close requires a journaling input, but that input can set
`required: false`; the returned completion policy then counts only the anchor.
This is an implemented discrepancy between the rule's required-kind label and
the emitted step requirement.

The duration budget removes tail steps whose kinds are neither anchors nor
rule-required. It does not fail when the protected core exceeds the requested
maximum, and it can remove an optional-kind step even when that step carries
`required: true`. The displayed cases preserve those limits.

The feature builder accepts any nonblank modality string while the canonical
contract schema rejects values outside its modality taxonomy. Tone context is
selected from literal searchable terms with ordered precedence. Its evaluator
uses deterministic text heuristics; because the adapter supplies no render
duration, the pace axis records a neutral `0.5` and says the pace check was
skipped. Dharma policy resolution returns a separate member-tradition-consent
decision, but the tone-review record does not carry or enforce that decision.

## Design and story

Visual thesis: a restrained cream field manual where one morning ritual moves
through rule tables, native relationship diagrams, exact record specimens and a
visible publication gate. Terracotta marks decisions and refusals; typography
and spacing carry hierarchy without decorative panels.

Content plan: establish the three record planes, inspect all six moment rules,
assemble the morning case, test ingredient and duration failures, compare local,
feature and contract invariants, trace script fingerprinting and Lilith policy
selection, then close on the current-review gate and missing integration owner.

Interaction thesis: retain the existing reduced-motion-aware slide entrance,
source reveal and center navigation. No new interaction is required; all
relationships and refusal paths remain complete in standalone HTML, embedded
HTML and PDF.

## Acceptance and verification

- All twenty pinned source cases must pass and retain exact source hashes.
- The guide must build from canonical content, appear in the intended
  prerequisite order, and carry complete narration and a complete PDF.
- Every page must pass standalone and embedded desktop/mobile review; selected
  dense pages receive enlarged-text checks and material boundary nodes receive
  negative controls.
- The two reviewed handbook-unit assignments must bind exact headings, unit
  hashes, cited slide fingerprints and the current teaching-coverage report.
- Existing scripts, manifest entries, MP3 bytes and guide HTML files must remain
  exact against the baseline except for the new guide and derived catalog/center
  outputs.

No member, editor, tradition authority, rights holder or normal-speed listener
participates in this automated delivery. Those acceptance paths and the missing
application/converter ownership remain open in the full V1 program.

## Recovery and closeout

Draft rendering uses the ignored `.qa-draft-tara-ritual-assembly` fixture.
Narration is rendered sequentially to the canonical bank only after the draft
review passes. The guide enters the canonical catalog for embedded, audio and
PDF verification before publication. Stop only package-owned renderer/browser
processes and retain logs and valid completed tracks. Publication requires
focused and current whole-library checks, a dated receipt, normal hooks, a
task-scoped commit and the branch tip on `origin/main`.
