# Arete humane engagement · working package · 9 September 2026

This package completes the remaining meaningful units in
`V1/features/domain-arete.md`: the product surface map, contract family,
humane-streak policy, recovery engine, missed-day derivation, friction and
coaching vocabularies, weekly review, cross-domain links, BFF persistence,
Living Offerings, implementation status and related reading. The existing Arete
introduction already teaches the check-in contract, so this guide follows that
record through the actual web, mobile, service and local-store consumers.

Implementation baseline and primary citation pin:
`1ceae52c7896a59ac74a4429d410f3218f0b91c6`.

Visual thesis: present Arete as a calm accountability field guide rather than a
celebration dashboard. A continuous day-to-week trace moves across cream, ink
and terracotta diagrams. Explicit evidence labels distinguish contract, executed
model, local state, service authority and product fixture. Recovery branches
stay visually prominent so a missed day reads as a policy decision, not a moral
failure.

Content plan: open with one member day across the eleven declared consumer route
directories and explain why the weekly alias yields ten rendered destinations.
Establish the five evidence levels and the separate consumer and power-user
product trees. Inspect the check-in, habit, goal and routine contracts before
comparing the contract's humane-streak vocabulary with the domain recovery
engine and the BFF's aggregate streak fold. Follow one write through
authentication, PostgreSQL status projection, recalculation and the records that
are not persisted. Then compare the contract and domain friction taxonomies,
coaching recommendations, weekly-review schema and actual closure store. Finish
with cross-domain relationship builders, Living Offering safety and provenance,
a current implementation matrix and a worked day-to-week trace.

Interaction thesis: use the established deck navigation, narration, source
drawer and keyboard-accessible product-image inspection. State-path diagrams let
the reader reveal each boundary from check-in through recovery and review. Real
product crops anchor the hub, recovery, review and offering surfaces; each crop
carries its preview or unavailable-state context in the slide itself. All
claims, branches and source boundaries remain complete in static HTML and PDF.

## Source review and corrected boundaries

The handbook accurately describes the intended record family, check-in
invariants and broad product areas. The pinned implementation requires sharper
boundaries in six places:

- `registerAreteRoutes` mounts adapter reads, practice-home and coach-response
  routes. The consumer room, habits, check-in, streak, review and offering
  routes are implemented in `domain-stubs.ts`, despite that filename.
- The contract's humane policy uses skip/decline preservation, grace cadences,
  recovery-prompt counts and visual language. The recovery engine uses freeze
  allowance, grace days and milestone days. The BFF maps a stored habit into the
  engine's default policy, so no inspected write path preserves a per-habit
  canonical humane policy end to end.
- PostgreSQL check-ins persist one row per member, habit and date and project
  five statuses into engaged, grace and no-count states. Partial always projects
  to fifty percent, skip and decline omit the contract-required reason, and the
  projection still uses the house tenant. The current streak fold preserves all
  grace dates without applying the contract's per-habit flags or allowance.
- The recovery engine is implemented and tested, but it derives daily misses in
  UTC, treats missing or invalid completion timestamps as zero misses, and
  returns one of five engine stages. The inspected consumer check-in path does
  not persist canonical `MissedDay`, `StreakRecord`, `RecoveryRecord` or `Plan`
  rows; it persists check-ins and lean habit streak fields.
- The weekly-review contract has four authored sections and a lifecycle. The
  live read computes kept/rest highlights and curated prompts from check-ins.
  The customer review page displays a preview fixture, while its durable
  mutation records only the week label and `closed` or `saved-private` action.
- The Living Offering workspace now has a controlled intention textarea, so the
  handbook's hardcoded-intention statement is stale. Its scene timeline, score
  choices and sharing rules remain fixture data. Keep/send runs a Lilith safety
  check and stores a recipient, occasion, short intention preview and action;
  the receipt contains development hashes and a `dev-manifest-reference`, not
  rendered media bytes or a production C2PA signature.

The product trees also need evidence labels. `/arete` has eleven page
directories, but `/arete/weekly` redirects to `/arete/review`. Recovery, goal,
review, coaching, patterns and offering read fixtures; streak and habits can
fall back to fixtures; plan mixes a room read with authored fixture fields. The
depth banner labels most of these routes as previews. The separate
`/domains/arete/*` tools use a browser-local `oshun.arete` store with a smaller
shape, points and direct streak increments. Mobile displays a polished daily
check-in, but its completion handler currently changes local React state,
telemetry and haptics without posting a check-in.

Arete has real Tara, Nisaba, Metis and Veritas relationship builders and shared
concept-graph linkages. Those functions return typed deterministic relationships
for supplied state. They do not establish a consumer navigation or persistence
handoff unless the inspected caller performs one.

## Working story

1. Follow one member from a morning habit through a missed day and weekly
   review.
2. Separate contract, executed model, browser state, service state and preview
   fixture evidence.
3. Map the consumer route tree and the separate power-user tools.
4. Inspect check-in invariants, then habit, goal and routine record anatomy.
5. Compare the contract policy, recovery engine policy and BFF aggregate fold.
6. Execute the five recovery stages and identify the UTC and missing-history
   decisions.
7. Follow a check-in POST into PostgreSQL and account for every retained field.
8. Show which canonical recovery records stop at schema or model evidence.
9. Compare the contract and domain friction vocabularies and recommendation
   ranking.
10. Trace live review highlights, the preview review surface and durable closure
    records.
11. Map Tara accountability, Nisaba reading, Metis study, Veritas grounding and
    concept-graph relationships to their real return values.
12. Inspect Living Offering input, safety admission, persisted action and
    development receipt.
13. Resolve the opening case with a current implementation matrix and a concrete
    connection priority.

## Acceptance and recovery

The final guide must preserve all twenty-two remaining exact source-unit hashes
and exact slide fingerprints in its assignment. Focused characterization tests
must prove route ownership, consumer and power-user surface counts, contract
invariants, policy divergence, recovery branches, PostgreSQL projection,
aggregate streak behavior, mobile-local completion, friction vocabularies,
review durability, cross-domain relationships and the current Offering boundary.
Existing owning contract, domain, BFF, web and browser suites remain separate
evidence.

The candidate must pass the presentation builder, focused manuscript and source
contracts, assignment validation, narration integrity, standalone Playwright,
Presentation Center review, PDF text and bounds checks, and complete optical
review. Product-source captures must verify the unavailable or preview labels
they display. No local, source-level, fixture-backed or development-manifest
pass is described as production, deployed-service or human acceptance.

If generation is interrupted, stop only the owned renderer or browser process,
retain its log and resume missing narration by fingerprint. Canonical content is
promoted from the draft only after the rendered candidate passes. Existing guide
HTML and audio are hash-compared so a failed package cannot silently replace
previously delivered work.
