docs: define world-to-scene validation and portable export slice

[frankxai]

The existing world graph, generation tools and canon validators do not yet provide evidence of one revision-bound author workflow. The runtime validators also carry independent reference tables, so a successful lexical check cannot establish current canon authority.

Adds docs/execution/2026-09-14-alignment.md with inspected paths, a bounded implementation-agent prompt, acceptance cases and explicit ownership boundaries. Refs #125.

The brief reuses the existing monorepo and Worldbuilding OS. It distinguishes original author-world profiles from official Arcanea rules, preserves locked canon and creator rights, and requires a scene export to retain source, rule, rights and review evidence.

Existing work remains owned by #98, #103 and PRs #90/#104/#116/#122/#123. This brief does not resolve their canon rulings or copy new lore into the product.

Validation: read repository instructions, licensing/worldbuilding documents, graph/generator/validator/MCP source and open issue/PR context. Reviewed for scope and private-data minimization. One documentation file; no runtime or browser tests, provider use, deployment, canon edits or promotion. This PR does not close the implementation issue.

[claude]

Review: docs: define world-to-scene validation and portable export slice

This is a documentation-only change (one new file, docs/execution/2026-09-14-alignment.md, +175/-0). Verified against the current main tree and open issue/PR tracker rather than reviewed as code.

Verification

• All 10 referenced source paths exist on main as cited (docs/worldbuilding/SYSTEM.md, docs/worldbuilding/README.md, packages/arcanea-mcp/src/tools/{creation-graph,generators,validate}.ts, packages/os/src/canon-validator.ts, packages/arcanea-mcp/src/index.ts, .claude/ci/lore-lint.mjs, LICENSING.md, AGENTS.md).
• All 8 referenced issues/PRs (Bind author scene exports to versioned world rules and validation evidence #125, Repo-wide superseded-name sweep: Thessara→Vaelith, Amaterasu→Source #98, World Genome v0 + non-canon Future Branch simulator on existing lore architecture #103, Govern Arcanea canon, rights, inventory, and ARC-REL-001 #116, canon: separate Starlight worldview from Arcanea fiction #122, Separate Starlight ontology from Arcanea sacred mythology #123, [codex] Launch the Living Arc transmedia story system #90, Add STAGING World Graph + Metrics + Starlight Corps Emblem #104) exist, are open, and their titles are consistent with how this brief characterizes their scope (no misattribution found).
• File placement follows this repo's own convention (/docs for documentation) and does not touch the locked canon vault or any runtime code.

Strengths

• Explicitly scopes itself as "documentation only" and lists what it does not do (no canon edits, no promotion, no runtime/browser tests, does not close Bind author scene exports to versioned world rules and validation evidence #125) - good practice in a repo where spec/canon drift has caused real incidents (issue Repo-wide superseded-name sweep: Thessara→Vaelith, Amaterasu→Source #98).
• The duplicate-work table proactively disambiguates ownership against six other in-flight issues/PRs, reducing the risk of this brief becoming another source of drift.
• Correctly defers to lore-release-gate and web-release-gate for any lore/UI work spawned from this brief, matching the repo's load-bearing gates in .claude/CLAUDE.md.
• The acceptance matrix includes adversarial cases (foreign workspace/path traversal, provider failure, stale review after edits, missing rights), not just happy-path criteria - appropriate rigor for a spec handed to an implementation agent.

Concerns / suggestions

1. No enforcement mechanism for the "no code" boundary. The doc repeatedly asserts scope limits (do not touch canon, do not build a new platform, do not duplicate Repo-wide superseded-name sweep: Thessara→Vaelith, Amaterasu→Source #98/World Genome v0 + non-canon Future Branch simulator on existing lore architecture #103/Govern Arcanea canon, rights, inventory, and ARC-REL-001 #116) as prose only. Given this repo already has automated ratchets for lore drift (lore-lint.mjs) and web guidelines, consider whether acceptance-matrix items touching canon/rights (e.g. "Locked-vault preservation: byte-identical before/after") should eventually be backed by a checked assertion rather than relying on the implementation agent to honor the prose.
2. The audit snapshot pins to one main SHA (fc1d1c0...) that will drift quickly given this repo's velocity (10+ open PRs touching adjacent surfaces). Since the disposition table is itself evidence other agents will rely on, consider an explicit "re-run the audit if stale" trigger beyond the single sentence currently there.
3. Minor: "the existing naming sweep owns their corpus-wide repair" (referring to Repo-wide superseded-name sweep: Thessara→Vaelith, Amaterasu→Source #98) reads clearly here but stands somewhat alone for a fresh implementation agent who hasn't seen Repo-wide superseded-name sweep: Thessara→Vaelith, Amaterasu→Source #98's discussion - could use one more clause of context.

Not applicable to this PR

• Code quality/bug/performance/security review: no source code changed.
• Test coverage: none expected for a planning document; the brief itself correctly identifies runtime tests as out of scope for this slice.

Overall: well-scoped, internally consistent, and its factual claims about the codebase check out. No blocking issues from a documentation-review standpoint.