[nynexman4464]

Stack

1. docs(core): define the internationalization contract #6048 — internationalization architecture and provider contract
2. docs: define RTL audit evidence policy #6046 — RTL audit evidence policy
3. docs: define bidirectionality design intent #6047 — bidirectionality design intent

Intent

Define the evidence boundary between a story-level D1 fixture exclusion and component-wide RTL applicability. This record specifies intended behavior; the fixture metadata implementation is separate in #7489 and is not part of this stack.

Authority

The record is promoted to authority: current with approved_by: nynexman4464 and approved_at: 2026-10-15, on the owner's approval of the reviewed claims. The exact-head /approve-spec command on this pull request records that decision for the final head. Approval does not ship INV12–INV16 enforcement: the record keeps that implementation state explicit, and component-wide verified N/A, historical debt, D7 size enforcement, and required-check rollout remain open decisions.

Existing authority searched

• Canonical id: architecture:rtl-audit-evidence; reuse the existing docs: define RTL audit evidence policy #6046 record. Current architecture:knowledge-contracts separates owner decisions, claim-scoped authority, and evidence; AST-030 owns CI routing rather than RTL applicability semantics. AST-020 retains target-size ownership.
• Paths/symbols: apps/storybook/rtl-audit/, Storybook preview metadata, SVGIconRegistry, scripts/component-packages.cjs, pr-rtl scope/shard/join, weekly summaries, and physical-property lint; rtlAudit.D1, candidate ordering, coverage rollup, verified-N/A declarations.
• Semantic terms: RTL, bidirectional, directional icons, caller-supplied glyph data, mirror, applicability, verified N/A, coverage gap, D1/D5/D6, planned/completed identity.
• Refreshed open overlaps: test(rtl): make fixture applicability order-independent #7489 is the fixture-policy implementation; test(rtl): close applicability gaps with fail-closed evidence #5885 proposes component-wide byte-bound verified-N/A admission. These are separate boundaries. This draft does not adopt test(rtl): close applicability gaps with fail-closed evidence #5885's whole-component closure policy or recreate the superseded fix(rtl): baseline pre-existing applicability gaps #5988 debt baseline. docs: define bidirectionality design intent #6047 supplies human RTL meaning; docs(core): define the internationalization contract #6048 supplies the provider/DOM direction boundary.

Authority and ownership

Extend architecture:rtl-audit-evidence, not a new parallel policy record. nynexman4464 is listed first, retaining cixzhang and imdreamrunner. Human RTL meaning remains with design:bidirectionality; provider behavior remains with the internationalization record.

Cross-cutting evidence

The global testing route supplies selected RTL evidence and exemption-review claims without importing product intent or workflow topology into every component. design:bidirectionality owns the shared component-facing RTL outcomes; local records own component states and verification. The route does not settle component-wide N/A, debt, or gate rollout.

Policy clarification

• D1 checks directional controls and cues, not the raw geometry of caller-supplied reference glyphs displayed as data.
• A reference fixture may declare D1 non-applicable with typed resolved metadata and a nonempty semantic rationale.
• Missing rendered metadata, malformed declarations, unsupported dimensions, or missing reasons fail closed.
• D5 and D6 continue scanning every in-scope story. A D1 exclusion neither marks a component measured nor grants component-wide verified N/A.
• Adding or widening an explicit D1 exemption, or changing an exempt story's semantic role, requires normal human PR review. No special approval is required for automatic no-icon N/A or every unchanged audit run.
• Every declared exemption in the audited scope must appear in the review-facing summary with component, story, dimension, and rationale, even if D1 selected another story.
• D1 candidates use source path, source export order, and story id. Planned/completed identities remain stable when an excluded fixture causes another candidate to run.

Component-wide verified-N/A closure, historical coverage-debt admission, required-check rollout, and D7 size enforcement remain separately identified gaps/decisions. This change does not decide them or require them to be settled before claim-scoped review of the fixture policy.

Implementation relationship

The base implementation does not contain the new D1 metadata transport or exclusion tests. The record says so explicitly. #7489 provides candidate implementation evidence: focused tests, reordered-story before/after proof, fail-closed metadata handling, and a registry-only run showing continued D5/D6 coverage. Those results do not independently approve the policy.

The complete exclusion inventory and semantic review requirements in INV16 are intended requirements, not already-implemented summary enforcement. They remain separate follow-up work; successful candidate metadata tests do not ratify an exemption's meaning.

Validation

• pnpm check:repo, knowledge validation, formatting, and git diff --check pass.
• Fresh delivery-branch validation passes all 152 selected knowledge/RTL join/completion/scope tests. The owner-validator prerequisite landed in fix(specs): use current repository approver membership #7538; all 121 knowledge tests pass on current main, and the promoted record validates. Earlier combined-stack raw-Intl boundary coverage also remains green.
• The fixture hardening has a negative null-metadata unit case and browser before/after proof for malformed metadata on a later unselected story. Valid metadata still runs D5 and D6 for both stories.
• The cross-cutting metadata passes all 100 knowledge-validator tests and focused synthetic route checks, including draft omission and selective claims; no underlying audit implementation or approval state changed.

Writing rubric results

R1 narration: replaced the draft-origin narrative and approval dependency on unrelated open decisions with the claim-scoped evidence boundary. Promotion removed draft-readiness wording rather than narrating the approval.
R2 research: current scores and historical experiments are not normative record content.
R3 evidence: the D7 gap describes the missing size assertion and observable failure, not the old PR that demonstrated it.
R4 rejected alternatives: no rejected-options narrative was added.
R5 change references: removed the superseded debt-baseline PR and historical touch-target PR from the record.
R6 open questions: component-wide N/A, debt, and rollout questions remain open; the fixture-only policy does not answer them.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
astryx	Ready	Preview	Oct 15, 2026 6:05pm UTC

[github-actions]

PR Analysis Report

📚 Storybook Preview

View Storybook for this PR

🧪 Sandbox Preview

View Sandbox for this PR

No new or modified components detected.

Bundle Size Summary

Package	Size (ESM)	Size (CJS)	Gzipped
@astryxdesign/core	N/A	4.9KB	1.2KB

Accessibility Audit

Status: No accessibility violations detected.

Visual Regression

38 of 474 shot(s) changed.

A change here is a question, not a failure: check whether the after is the
picture you intended. Record that review on the PR. Baseline maintenance is an
explicit dispatch of CI; this report never rewrites the baseline or adds a release gate.

component	story	theme	mode	pixels
Selector	Indicator Space Bottom Sheet Narrow End RTL	neutral	light	754,077
Selector	Indicator Space Bottom Sheet Wide Start	neutral	light	740,188
Chat	Full AI Chat	probe	light	238,087
Chat	Full AI Chat	probe	dark	237,060
Collapsible	Dividers	probe	dark	66,445
Collapsible	Dividers	probe	light	66,445
Chat	Default	neutral	dark	33,205
Chat	Default	neutral	light	32,876
Chat	Empty State	probe	light	23,531
Chat	Empty State	probe	dark	22,622
ChatAutoScroll	Scroll Behavior Comparison	neutral	dark	15,378
ChatAutoScroll	Scroll Behavior Comparison	neutral	light	14,107
Chat	With Attachments	probe	light	13,406
Chat	With Attachments	probe	dark	11,232
ProgressBar	With Target Mark	probe	dark	7,895
ProgressBar	With Target Mark	probe	light	7,895
TableRowExpansion	Detail Panel	neutral	dark	5,150
TableRowExpansion	Detail Panel	neutral	light	4,068
SideNav	Resizable in AppShell	probe	dark	3,846
SideNav	Resizable in AppShell	probe	light	3,846

and 18 more.

---

_{Generated by PR Enrichment workflow | Storybook | Sandbox | View full report}

[cixzhang]

/approve-spec a1a63a9

[cixzhang]

Thanks for separating measured RTL evidence, verified non-applicability, historical coverage debt, and merge-gate policy instead of treating one implementation as already decided. Approved as the canonical RTL audit-evidence record at exact head a1a63a9238fe5d382cb49b3b3c493b6d48e83ed3.

[astracat-bot]

AI review status for this pull request.

Review status	Updated
🔄 In progress (for maintainers only)	Oct 15, 2026, 6:07 PM UTC

[nynexman4464]

/approve-spec 7355ffd