core/docs/handoff/core-logos-studio-LG2-brief-2026-06-14.md

LG-2 — CORE-Logos Studio: route shell + Overview / Identity / Safety

Date: 2026-06-14 Wave: read-only CORE-Logos Studio (L1–L5). Parent pack: docs/handoff/core-logos-studio-readonly-briefs-2026-06-14.md. Plan: docs/workbench/core-logos-studio-plan.md. Depends on: LG-1 merged (workbench/logos.py + /logos/* endpoints + the 10 Logos* schemas + their NOT_YET_MIRRORED allowlist entries). Do not start until LG-1 is on origin/main; rebase on it. Scope (hard): route shell + Overview / Identity / Safety tabs only. Lexicon/Glosses/Morphology (LG-3) and Alignment (LG-4) are separate follow-on briefs — do not implement their tabs here. The /logos/contents and /logos/alignment endpoints exist but LG-2 reads only /logos/packs, /logos/packs/{id}, and /logos/packs/{id}/safety.

Non-negotiables (Wave M doctrine)

• No theater, no recompute. Render only what the live /logos/* endpoints return. The workbench computes nothing the engine owns.
• Read-only. No mutation calls; no "Draft proposal" affordance this wave.
• Determinism. No nondeterministic layout/motion. Golden-file/render tests.
• Doctrine gates green: route conformance, schema drift, enum coverage.

Build

1. Route registration (src/app/routes.ts)

• Add /logos to WORKBENCH_ROUTES in the Substrate section (alongside Packs / CORE-Logos), wired through ROUTE_ELEMENTS as a lazy route element (Phase-E route-chunk pattern — match the existing lazy entries).
• Honest keyboard model: assign a ⌘-digit only if one is free; otherwise palette-only (match the existing 10-pin-+-palette pattern). Bump the nav/palette/guide counts the registry derives.
• routes.test.tsx must stay green: element-map parity, unique ids/paths/ digits, palette reachability, landing coverage.

2. TS schema mirror (src/types/api.ts + drift gate)

• Add TS interfaces for the 10 Logos* schemas (LogosPackSummary, LogosPackOverview, LogosPackContents, LogosLexiconRow, LogosGlossRow, LogosMorphologyRow, LogosAlignmentRow, LogosMorphologyLinkIssue, LogosAlignmentTargetIssue, LogosSafetyReport) — field-faithful to workbench/schemas.py / workbench-ui/schema-snapshot.json.
• Shrink NOT_YET_MIRRORED in src/design/doctrine/schemaDrift.test.ts: remove each Logos* entry as its interface lands. The gate is shrink-only — a class gaining an interface while still listed fails. (Mirror all 10 even though LG-2 only renders a subset; the contract is per-schema, and LG-3/LG-4 consume the rest.)

3. SafetyVerdict UI mapping + coverage (where it is rendered)

• Add SafetyVerdict to src/design/components/badges/types.ts and a safetyVerdictMeta (satisfies BadgeMeta<SafetyVerdict>) to mappings.ts, covering all four values: clear / warning / failed / unknown — with honest meanings (unknown ≠ clear, warning ≠ clear).
• Extend scripts/dump-enums.py to source SafetyVerdict from workbench/schemas.py (now justified — it is actually rendered), regenerate workbench-ui/enum-snapshot.json, and add a SafetyVerdict assertion to enumCoverage.test.ts (exact-coverage vs safetyVerdictMeta). This discharges the deferral recorded in LG-1.

4. /logos page shell

• Layout: SplitPane — Pack Universe rail · Studio Workspace · Evidence Inspector (reuse the existing Panel/TabBar/SplitPane primitives).
• Pack Universe rail: GET /logos/packs; group by role (depth_root / depth_relation / logos-cognition); count badges; a SafetyVerdict badge per pack (text label, not color-only).
• Bottom status strip (persistent): selected pack · checksum status · gate/OOV · proposal mode: none — read-only.

5. Tabs (this wave)

• Overview — GET /logos/packs/{id}: tri-language role framing + counts + the safety verdict badge. Holonomy: render holonomy_case_count (0) as missing_evidence — no tab, no proof card, no success state.
• Identity — manifest passport fields + raw manifest via the existing StableJsonViewer.
• Safety — GET /logos/packs/{id}/safety (LogosSafetyReport): checksum status, domain-contract status, dangling morphology links, invalid alignment targets (LG-1 surfaces the undeclared en-collapse-* anchors here — show them), epistemic-status counts, missing_holonomy_refs as unknown, known gaps, verdict. unknown/warning must never render as clear.

6. Evidence subject + address grammar (src/app/evidenceAddress.ts)

• Add the logos_pack subject kind (alongside pack / run / vault_entry / audit_event): union member, equality, encode/parse.
• Address grammar: logos:<pack_id> (distinct from the existing pack:<id>).
• Right-inspector projection + copyable pointer (Cmd+Shift+C) + ?inspect= URL-sync, matching the existing subject pattern.
• Extend evidenceAddress.test.ts with a logos_pack round-trip case.
• Defer logos_entry / logos_gloss / logos_morphology / logos_alignment_edge subject kinds to LG-3/LG-4 (vocab-commitment: add a kind only when its tab renders it).

Acceptance / tests

• routes.test.tsx — /logos registered, contract parity, counts bumped.
• schemaDrift.test.ts — 10 Logos* interfaces present; NOT_YET_MIRRORED shrunk by 10; gate green.
• enumCoverage.test.ts — SafetyVerdict exact-coverage green.
• evidenceAddress.test.ts — logos_pack round-trip.
• Render tests: Overview / Identity / Safety against fixture payloads, including the honest absent-holonomy state (count 0 + missing_evidence, no proof card) and a warning/unknown verdict never shown as clear.
• pnpm build clean (route lazy-loaded, no chunk-size regression).

Out of scope (separate follow-on briefs)

• LG-3 — Lexicon / Glosses / Morphology tabs + their evidence subjects.
• LG-4 — Alignment tab (deterministic DAG centerpiece, invalid-target warnings) + holonomy-absent confirmation.

Wrinkles to carry

• The LG-1 hazard fix means real logos packs read WARNING (e.g. grc_logos_cognition_v1 has 5 undeclared en-collapse-* targets) — that is correct and should be visible in Safety, not smoothed over.
• CLEAR is unreachable as an overall verdict by design while holonomy proof is absent (LG-1: a logos pack tops out at UNKNOWN). Don't add UI that implies a clean pack should be green — unknown is the honest ceiling until W-Holonomy.
• speculative is the expected default epistemic_status (ADR-0021) — label it neutrally, not as a defect.