# Claude · Crown Key tooling assignment

R01 · Updated 2026-10-17

[Current five-seat plan · R02](https://thelostsouls.world/CrownKey/production/plan/) · [R02 AI start here](https://github.com/215GrandAve/lost-souls/blob/main/docs/crown-key-production/R02/AI_START_HERE.md)

The current five-seat production plan is at https://thelostsouls.world/CrownKey/production/plan/ and docs/crown-key-production/R02/AI_START_HERE.md. Three OpenAI seats and two Claude seats work in parallel: Quik starts Crown Landmark authoring/staging while the team completes Split Complex and Terrace Tower packages. Every enclosed building receives a lobby plus one or two useful interior spaces, with Atlas scene requirements taking priority. Open structures receive appropriate arrival and access spaces. R01 remains technical and historical support.

Read current repo instructions and use a task-specific branch/worktree with exact claims. PRs #45 and #51 are merged; retain their source history and check current HEAD before new changes.

## Concrete input handoff

The disposable R02 sample mapping and sealed TEST-ONLY package are merged. Earlier CLAUDE_NEXT.md review instructions are dated history. Use the merged tooling and reconcile the newer Split H1 scope with the old Candidate B structure before assigning proposed production values. Claude contract-pilot PR #51 is merged into main (merge 7a81eb8, head 3701c09), including reviewed R02 mapping, deterministic drawings/blockouts/cameras and the TEST-ONLY Quik import package. Manifest SHA-256 a232289bb781ada7fed5718b8b660a2834e4cc471ac2d8dc6ddeecbdfe578aac. Quik’s actual Unreal results remain pending. Both test contracts remain refused by the production issue gate; automatic production packaging integration is still a separate task.

## Your ownership

Own dimension/ledger schema, validators, migration proposals, SVG drawing generators and Blender blockout export. Codex owns plan/ledger entries and proposed asset/sample values. Ask Codex for specific missing design inputs; do not silently author values in another owner's contract.

## Reviewable checkpoints

### A · Baseline and schema proposal

Baseline SHA, active PR inventory, exact branch/path claims, schema changes and migration impact. PR #51 (merged 7a81eb8) published schema/path/migration evidence after Codex review.

One owner per edited file; extend existing contracts without silently modifying released R05 sources.

### B · Dated decision sheets

Schematic bridge, clearance-datum/route and shoreline options, affected assets, sources and a separately marked recommendation.

Choices unblock affected site work only after Legend records a selection. Rendered concept art comes from the visual team.

### C · Synthetic fixture pipeline

SVG drawings and Blender blockout from the same test contract; known-size asymmetric orientation fixture. PR #51 (merged 7a81eb8) supplies reviewed fixture evidence and the TEST-ONLY Quik import package; Quik’s Unreal results are pending and this is not engine acceptance.

Check value changes, openings/markers/reservations, unknowns, revision stamps and test-only enforcement. Record actual tool versions.

## Geometry and engine feedback

Help Quik run the merged fixture/sample and record actual import/traversal feedback while assigned asset owners advance Split Complex and Terrace Tower. Use their proposed contracts for deterministic drawings, editable blockouts and saved cameras. Correct affected outputs when engine evidence arrives; independent design/modeling work need not wait. Do not treat upper-floor room-fit results as evidence of entrance traversal.

## Required engineering behavior

- Preserve Building R05 A-E tiers. Image ranges are planning allowances, not quotas; a canopy and major story landmark need different coverage.
- Codex proposes initial sample and Split Complex design values, explicitly marked proposed/test-only as applicable. Never present inferred image dimensions as measured.
- Real openings, visual markers and space reservations have different geometry and acceptance behavior.
- Issue validation and Verify reject test-only contracts as production releases. The merged tools provide the gate; tools/build_quik_handoffs.py still needs automatic integration. Proposed production values can support a modeling issue when essential geometry is complete; issue status does not confer final acceptance.
- Per-asset contracts hold dimensions; the plan/ledger holds references and readiness. Derived copies must retain revision/hash provenance.
- Choose interchange format after documenting actual Blender/Unreal importer versions and known-size/orientation tests; recommend with reasons and let Quik validate.
- Schema version changes and asset value changes have separate owners. No generator may silently overwrite authored values or migrate all assets before review.
- No silent changes to district geometry, source artwork, issued IDs, accepted references, private credentials or unrelated .com pages.

Use one dimension contract per asset and the merged asymmetric known-size fixture, explicit units/axes/import settings and revision/hash stamps. Retain checks for dimensions, real openings versus markers/reservations, missing values and test-only release rejection. The validator and Verify controls are implemented in `tools/contracts/`; automatic integration with `tools/build_quik_handoffs.py` remains outstanding.

Recommend interchange format with reasons and actual importer constraints; Quik confirms through his test. Record the actual tool versions and whether headless Blender executed. Add a short README with generator/validator commands so other agents can reproduce results. Do not silently change a published protocol or batch-migrate all contracts before review.

## Return and publication

Return source revisions, ownership/path claims, a schema/migration diff, schematic decision sheets, fixture outputs, validations, limits and a focused PR when authorized. Checkpoints should remain reviewable even if a session ends early.

Legend authorizes Claude to work and push on his own branch without repeated approval. Legend also explicitly authorizes main publication within the current Crown Key pilot/preparation phase, including PR #50 and its ordinary publication/status fixes, without asking again. Run the required checks. Preserve exact path ownership and the separate design, asset, fixture and engine acceptance gates.
