stylegallery 0.1.0

package/README.md

@@ -0,0 +1,313 @@
+---
+type: Repository Guide
+title: StyleGallery
+description: Governed gallery of portable interface knowledge organized by domain.
+---
+
+# StyleGallery
+
+StyleGallery is a governed gallery of portable interface knowledge. It separates reusable spatial patterns, product-layer motion guidance, design-engineering practice, and platform-specific references into explicit domains with different evidence and ownership boundaries.
+
+Primary role: repository guide.
+
+The existing Layout corpus remains a gallery of minimal, portable CSS layout patterns at its current paths. Each pattern documents one primary spatial problem and the smallest robust HTML/CSS structure that solves it. Motion, visual treatment, and platform guidance do not expand reusable Layout pattern CSS; they live in their own domains and carry explicit evidence boundaries.
+
+[Consumer Reference](consumer-reference/index.md) is shared non-domain infrastructure for optional consumer-owned reference handoffs. It carries schema, routing, provenance, and evidence metadata without owning profiles, visual values, components, or a sixth domain.
+
+[Agent-Native StyleGallery](consumer-reference/agent-native/README.md) is the machine-facing entry point over that governed knowledge. It provides content-addressed identity, claim/evidence/governance records, deterministic retrieval, a shared operation registry, the `sg` CLI, and a read-only MCP projection. It does not create a sixth domain, replace the Markdown corpus, or permit MCP clients to mutate governed knowledge.
+
+## Quick Start
+
+StyleGallery requires Node.js 22 or newer. Run a command without installing anything globally:
+
+```sh
+npx stylegallery discover --format json
+```
+
+Or install the CLI globally:
+
+```sh
+npm install --global stylegallery
+sg discover --format json
+```
+
+Common read-only commands:
+
+```sh
+sg resolve sg:profile/editorial-reference-profile --format json
+sg claims sg:profile/editorial-reference-profile --format json
+sg context sg:profile/editorial-reference-profile --format json
+sg ops --format json
+```
+
+Every command writes deterministic JSON to stdout. Invalid input returns an error object and a nonzero exit status. See [Agent-Native StyleGallery](consumer-reference/agent-native/README.md) for the command contract, StableRef and VersionID model, MCP resources, and trust boundaries.
+
+### Read-only MCP server
+
+Launch the packaged stdio server with:
+
+```sh
+npx --package stylegallery stylegallery-mcp
+```
+
+Example MCP client configuration:
+
+```json
+{
+  "mcpServers": {
+    "stylegallery": {
+      "command": "npx",
+      "args": ["--yes", "--package", "stylegallery", "stylegallery-mcp"]
+    }
+  }
+}
+```
+
+The MCP surface exposes governed read operations only. It cannot modify repository knowledge.
+
+## 한국어 빠른 시작
+
+StyleGallery는 Node.js 22 이상에서 실행됩니다. 전역 설치 없이 바로 사용하려면 다음 명령을 실행하세요.
+
+```sh
+npx stylegallery discover --format json
+```
+
+자주 사용한다면 전역으로 설치할 수 있습니다.
+
+```sh
+npm install --global stylegallery
+sg discover --format json
+```
+
+`discover`는 사용 가능한 인터페이스를 보여주고, `resolve`는 하나의 레코드를 조회하며, `claims`는 관련 주장과 근거를 분리해서 보여줍니다. `context`는 에이전트에 전달할 수 있는 제한된 컨텍스트 패키지를 만들고, `ops`는 지원하는 작업 목록을 반환합니다. 모든 결과는 자동화에 바로 사용할 수 있는 JSON입니다.
+
+CLI와 MCP의 상세 사용법은 [Agent-Native StyleGallery 가이드](consumer-reference/agent-native/README.md)를 참고하세요. 사람이 문서를 탐색하려면 아래의 도메인 표에서 목적에 맞는 진입점을 선택하면 됩니다.
+
+## Domains
+
+| Domain | Owns | Does not own |
+| --- | --- | --- |
+| [Layout](layout/index.md) | Semantic spatial structure, flow, sizing, alignment, containment, scrolling, and composition. | Brand, typography, color, shadow, animation, and product decoration. |
+| [Motion](motion/index.md) | Motion terminology, review procedure, and evidence-bounded practice guidance. | Universal timing/easing rules or permission to add motion to reusable Layout CSS. |
+| [Design Engineering](design-engineering/index.md) | Product-layer craft decisions and verification questions. | A second universal principle set or taste as evidence. |
+| [Game UI](game-ui/index.md) | Game-interface classification, hierarchy, reference records, and engine-specific implementation guides. | Reusable Layout CSS or claims that one engine structure is universal. |
+| [Platform Guides](platform-guides/index.md) | Bounded comparison with named platform conventions. | Affiliation, imitation, or authority over web and accessibility contracts. |
+
+The canonical domain manifest and provenance policy are in [StyleGallery Domains](DOMAINS.md).
+
+## Repository Entry Roles
+
+Use each root hub for one primary job.
+
+| Entry | Primary role | Use when |
+| --- | --- | --- |
+| [README](README.md) | Repository guide | You need the library purpose, policies, and task routes. |
+| [OKF index](index.md) | OKF bundle map | You need a compact knowledge-bundle table of contents. |
+| [Layout Planning Guide](GUIDE.md) | Planning workflow | You need to classify a screen before choosing patterns. |
+| [Layout Pattern Catalog](CATALOG.md) | Pattern lookup | You already know the spatial problem or pattern name. |
+| [Governance, Lifecycle, And Docs-As-Code](GOVERNANCE.md) | Governance reference | You need the source of truth, lifecycle, generated-file, ownership, or stale-audit rule. |
+| [StyleGallery Domains](DOMAINS.md) | Domain manifest | You need domain ownership, scope, lifecycle, page membership, or provenance. |
+| [Consumer Reference](consumer-reference/index.md) | Shared infrastructure contract | You need to declare a consumer-owned record or explain why one is not applicable. |
+| [Agent-Native StyleGallery](consumer-reference/agent-native/README.md) | Machine interface guide | A person or agent needs to discover, resolve, retrieve, or inspect governed StyleGallery knowledge through CLI or MCP. |
+| [Layout](layout/index.md) | Layout domain hub | You need reusable spatial patterns, recipes, or planning routes. |
+| [Motion](motion/index.md) | Motion domain hub | You need motion terminology, review procedure, or practice evidence. |
+| [Design Engineering](design-engineering/index.md) | Design Engineering domain hub | You need product-level interface-craft decision guidance. |
+| [Game UI](game-ui/index.md) | Game UI domain hub | You need to classify a game interface or understand its screen hierarchy. |
+| [Platform Guides](platform-guides/index.md) | Platform Guides domain hub | You need a bounded platform comparison. |
+
+## Task Routes
+
+Each common task has one primary route. Use secondary links only after the primary route answers the first decision.
+
+| Task | Primary route | Why |
+| --- | --- | --- |
+| `choose a StyleGallery domain` | [StyleGallery Domains](DOMAINS.md) | It separates domain ownership before a reader applies domain-local guidance. |
+| `browse reusable spatial guidance` | [Layout](layout/index.md) | It preserves the existing pattern, recipe, and planning routes. |
+| `name or review interface motion` | [Motion](motion/index.md) | It routes to bounded terminology and review guidance. |
+| `review product-level interface craft` | [Design Engineering](design-engineering/index.md) | It separates practitioner heuristics from shared quality gates. |
+| `compare adversarial consumer identities` | [Reference Profiles](design-engineering/reference-profiles/index.md) | It keeps non-default product values in related Design Engineering examples over one pinned Layout source. |
+| `classify a game interface or map it to an engine` | [Game UI](game-ui/index.md) | It separates engine-neutral roles from implementation-specific guidance. |
+| `compare a named platform convention` | [Platform Guides](platform-guides/index.md) | It requires platform and evidence boundaries before adaptation. |
+| `turn raw content into a homepage or ordinary webpage` | [Webpage Generation Workflow](guides/webpage-generation-workflow.md) | It starts with use case, content-to-layout fit, harmony, and handoff. |
+| `plan a screen before the layout problem is obvious` | [Layout Planning Guide](GUIDE.md) | It sequences task, content, scroll, recipe, and verification choices. |
+| `choose a pattern when the name is unknown` | [Decision Tree](guides/decision-tree.md) | It routes from constraints to pattern categories. |
+| `fill in requirements before selecting a pattern stack` | [Layout Brief Template](guides/layout-brief.md) | It captures content, constraints, and verification inputs. |
+| `stabilize repository terminology` | [Controlled vocabulary](guides/vocabulary.md) | It defines canonical terms, aliases, deprecated terms, and scannability rules. |
+| `compose a full screen from primitives` | [Layout Recipes](recipes/index.md) | Recipes map screen models to pattern stacks. |
+| `inspect which primitives a recipe depends on` | [Primitive To Recipe Matrix](recipes/primitive-to-recipe-matrix.md) | It names essential, helper, and substitutable slots. |
+| `look up a known layout primitive` | [Layout Pattern Catalog](CATALOG.md) | It is the generated pattern lookup surface. |
+| `browse pattern categories` | [Pattern Categories](patterns/index.md) | It groups generated patterns by spatial family. |
+| `check whether a layout or design claim is admissible` | [Quality Gates](quality/index.md) | It routes claims to gates and evidence boundaries. |
+| `prove repository checks and evidence coverage` | [Executable Evidence Coverage](quality/evidence/executable-evidence.md) | It maps validators, fixtures, CI commands, and their boundaries. |
+| `declare consumer reference applicability` | [Consumer Reference](consumer-reference/index.md) | It provides the required handoff field without moving consumer values into Layout. |
+| `use StyleGallery from an agent or automation` | [Agent-Native StyleGallery](consumer-reference/agent-native/README.md) | It explains StableRef resolution, CLI commands, read-only MCP resources/tools, receipts, and mutation boundaries. |
+| `prove an existing consumer migration` | [Consumer Migration Readiness](design-engineering/consumer-migration-readiness.md) | It requires thirteen explicit behavior classifications, runtime proof, adoption mappings, and source-bound page evidence when applicable. |
+| `change generated patterns, catalog, or governance policy` | [Governance, Lifecycle, And Docs-As-Code](GOVERNANCE.md) | It identifies source files, generated artifacts, validators, lifecycle state, and review ownership. |
+| `run findability QA` | [Tree-Test Findability QA](quality/index.md#tree-test-findability-qa) | It tests whether task routes are discoverable, not just linked. |
+
+## Link Policy
+
+- Navigation links move a reader to the next decision point in the repository. Root hubs, indexes, parent links, and next-step links are navigation links.
+- Citation links identify source lineage or evidence boundaries. They support a claim but should not be the only way to continue a task.
+- Dependency links identify generated, validation, or composition relationships. They explain what must stay in sync, such as `scripts/pattern-data.mjs`, generated pattern files, catalog entries, and validator fixtures.
+
+## How To Use This Repository
+
+- Start with [StyleGallery Domains](DOMAINS.md) when the owning domain is not already clear.
+- Use [Layout](layout/index.md), [Motion](motion/index.md), [Design Engineering](design-engineering/index.md), [Game UI](game-ui/index.md), or [Platform Guides](platform-guides/index.md) as the domain-local entry point.
+- Start with [Layout Planning Guide](GUIDE.md) when you are designing a screen before a layout problem is obvious.
+- Use the [Webpage Generation Workflow](guides/webpage-generation-workflow.md) when raw content needs to become a homepage or ordinary webpage before a layout recipe is obvious.
+- Use the [Documentation Mode Taxonomy](guides/documentation-mode-taxonomy.md) when adding or reviewing docs so each page has a clear primary reading mode.
+- Use the [Controlled vocabulary](guides/vocabulary.md) when a term affects routing, metadata, search, claim records, workflow handoff, or review decisions.
+- Use the [Decision Tree](guides/decision-tree.md) when you do not know the pattern name yet.
+- Fill out the [Layout Brief Template](guides/layout-brief.md) before choosing a pattern stack.
+- Use [Layout Recipes](recipes/index.md) when you need screen-level composition.
+- Use [Layout Pattern Catalog](CATALOG.md) when you already know the spatial problem.
+- Use [Quality Gates](quality/index.md) when a claim needs principle-backed evidence, visual QA boundaries, accessibility precedence, or design rationale.
+- Use [Consumer Reference](consumer-reference/index.md) when an implementation handoff must declare one repository-local JSON reference or a sentence explaining non-applicability.
+- Use [Agent-Native StyleGallery](consumer-reference/agent-native/README.md) when a person, script, or agent needs deterministic JSON discovery, StableRef/VersionID resolution, bounded context, operation metadata, or read-only MCP access.
+- Use [Consumer Migration Readiness](design-engineering/consumer-migration-readiness.md) only for a migration that declares a consumer-local conformance record; ordinary handoffs keep the existing `not_applicable` path.
+- Use [Governance, Lifecycle, And Docs-As-Code](GOVERNANCE.md) before changing generated artifacts, validators, lifecycle state, or ownership policy.
+
+## Layout Domain Principles
+
+1. Layout problems only
+   - Patterns exist to solve spatial problems, not to define a visual brand.
+   - Core CSS should stay focused on layout responsibilities.
+
+2. Semantic structure first
+   - Start with meaningful HTML before adding layout classes.
+   - Class names explain layout responsibility, but they do not replace landmarks, headings, lists, buttons, links, or form controls.
+   - DOM order, reading order, and focus order should remain logical.
+
+3. One primary spatial problem
+   - Each pattern should optimize one primary spatial problem.
+   - Secondary behavior is allowed when it is incidental or required for composition.
+   - If a pattern has multiple primary problems, split it into smaller primitives or mark it as a shell/composite.
+
+4. Minimum robust CSS
+   - Use the smallest set of declarations that keeps the pattern stable under real content and container constraints.
+   - Minimal does not mean fragile; constraints, overflow rules, intrinsic sizing, and fallbacks are valid when they carry declared responsibility.
+
+5. Portable plain HTML/CSS
+   - Plain HTML and CSS are the source of truth.
+   - Framework-specific versions are optional derivatives, not canonical patterns.
+
+6. Explicit constraints and change points
+   - Widths, heights, gaps, breakpoints, scroll containers, fixed/sticky anchors, and other constraints should be easy to find.
+   - Values users are expected to tune should be surfaced clearly.
+
+7. Named scroll ownership
+   - If a pattern scrolls, the scrolling element must be obvious.
+   - The pattern should make clear what scrolls, what stays fixed, and where height is determined.
+
+8. No decorative debt
+   - Decorative styling does not belong in reusable pattern CSS unless it is required to explain layout behavior.
+   - Demo-only visual aids should be separate from the reusable pattern.
+
+## CSS Authoring Policy
+
+- Prefer low-specificity, single-class selectors.
+- Avoid ID selectors for styling or layout naming.
+- Avoid deep combinators, selector chaining, and nesting unless the pattern is specifically demonstrating that tradeoff.
+- Use cascade layers when the CSS surface grows beyond isolated snippets.
+- Keep CSS declarations in alphabetical order inside each rule.
+- Prefer intrinsic sizing and content/container-driven adaptation.
+- Use viewport breakpoints only when the spatial problem is viewport-level.
+- Prefer container queries for component-local responsiveness.
+- Prefer logical properties for spacing and sizing unless physical direction is required.
+- Treat exceptions as explicit and local; use state hooks or `data-*` attributes for variants instead of escalating selector specificity.
+
+## Class Naming Policy
+
+Class names should read like a map of the layout structure.
+
+- Name layout responsibility, not appearance.
+- Make root-child, parent-child, area-element, and fixed-scroll relationships obvious.
+- Prefer pattern-scoped names for related nodes.
+- Avoid vague names like `container`, `wrapper`, `box`, `top`, `content`, and `bottom` unless the surrounding pattern name makes their responsibility unambiguous.
+- Avoid DOM-depth names such as `card_header_title_icon`; describe stable roles instead.
+- Do not use IDs for layout naming. IDs are reserved for JavaScript hooks or document-level targets when needed.
+
+Prefer:
+
+```html
+<div class="app_shell">
+    <header class="app_shell_header"></header>
+    <main class="app_shell_scroll_area"></main>
+    <footer class="app_shell_footer"></footer>
+</div>
+```
+
+Avoid:
+
+```html
+<div class="container">
+    <div class="top"></div>
+    <div class="content"></div>
+    <div class="bottom"></div>
+</div>
+```
+
+## Value And Token Policy
+
+- Tokenize reusable design intent, not every layout number.
+- Use tokens for stable, shared values such as `content-width`, `gutter`, `stack-gap`, `section-gap`, or density steps.
+- Keep browser/context mechanics in raw CSS: `auto`, percentages, `min-content`, `max-content`, `fit-content`, `clamp()`, viewport units, container-query units, and intrinsic sizing.
+- Put tokens behind logical CSS, for example `padding-block: var(--space-4)`.
+- Use breakpoint names for layout states, not device names.
+
+## Pattern Contract
+
+Use this shape when adding a new pattern:
+
+```txt
+Pattern name
+Category
+Primary spatial problem
+Secondary spatial problems
+When to use
+HTML structure
+CSS
+Core properties
+Properties that break the layout if removed
+Constraints and change points
+Scroll ownership
+Accessibility and source-order notes
+Browser/fallback notes
+Composition notes
+Anti-patterns
+```
+
+Suggested categories:
+
+- Containment
+- Centering
+- Stacking
+- In-line grouping
+- Split / Sidebar
+- Grid / Repetition
+- Viewport / Shell
+- Overlay / Exception
+- Media / Fit
+- Reveal / Density control
+
+## Verification Matrix
+
+Before accepting a pattern, verify the smallest relevant matrix:
+
+- Viewports: `320px`, `375px`, `768px`, `1024px`, `1440px`
+- Containers: tight, medium, roomy, and `100%` parent
+- Content: empty, short, long label, long paragraph, unbroken string
+- Direction: `ltr` and `rtl`
+- Writing mode: default, plus vertical writing mode if the pattern claims support
+- Interaction: default, hover, focus, active, expanded, scroll top/middle/bottom when relevant
+
+Acceptance checks:
+
+- The layout reflows without unusable two-dimensional scrolling.
+- Long and empty content do not break alignment or hide essential content.
+- Focus order remains logical and visible.
+- Scrollable regions and sticky elements behave as declared.
+- Screenshot diffs are reviewed when the pattern has visual fixtures.

package/consumer-reference/adapters/style-dictionary.config.mjs

@@ -0,0 +1,36 @@
+import path from "node:path";
+import StyleDictionary from "style-dictionary";
+
+export const adapter = Object.freeze({ name: "style-dictionary", version: "5.5.0" });
+
+const durationTransform = "stylegallery/duration-css";
+if (!Object.hasOwn(StyleDictionary.hooks.transforms, durationTransform)) {
+  StyleDictionary.registerTransform({
+    filter: (token) => token.$type === "duration",
+    name: durationTransform,
+    transform: (token) => `${token.$value.value}${token.$value.unit}`,
+    transitive: true,
+    type: "value",
+  });
+}
+
+export function createStyleDictionaryConfig(source, output) {
+  return {
+    log: { verbosity: "verbose", warnings: "warn" },
+    platforms: {
+      css: {
+        buildPath: `${path.dirname(output)}${path.sep}`,
+        files: [
+          {
+            destination: path.basename(output),
+            format: "css/variables",
+            options: { outputReferences: true },
+          },
+        ],
+        transforms: [...StyleDictionary.hooks.transformGroups.css, durationTransform],
+      },
+    },
+    source: [source],
+    usesDtcg: true,
+  };
+}

package/consumer-reference/agent-native/README.md

@@ -0,0 +1,157 @@
+---
+type: Interface Guide
+title: Agent-Native StyleGallery
+description: Machine-facing identity, query, execution, protocol-projection, and governed-learning contracts for StyleGallery.
+---
+
+# Agent-Native StyleGallery
+
+Primary role: machine interface guide.
+
+Agent-Native StyleGallery lets a person, script, or agent inspect the governed StyleGallery knowledge graph without learning repository paths first. The Markdown corpus remains the human-readable source material; the agent-native layer gives its governed fixture records stable identities, closed schemas, deterministic operations, and protocol-specific projections.
+
+This is shared non-domain infrastructure. It does not add a sixth StyleGallery domain, make example profiles canonical, or allow an MCP client to mutate repository knowledge.
+
+## Mental Model
+
+| Layer | Contract |
+| --- | --- |
+| Identity | A `StableRef` names a logical object. A `VersionID` binds that StableRef to canonical content bytes. Changing governed content creates a different VersionID. |
+| Knowledge | Claim, Evidence, Validation, Governance, and Policy records remain separate. Validation or governance never erases contradictory evidence. |
+| Operations | One registry defines operation names, schemas, effect class, idempotency, capability requirements, adapters, and `read_only` metadata. |
+| Execution | Task records preserve intent; Run records represent attempts; Effects record observable side effects; Receipts bind normalized inputs, outputs, policy decisions, and effects. |
+| Retrieval | View snapshots and context packages are deterministic, bounded, and provenance-linked. Retrieval does not assert truth or mutate the registry. |
+| Learning | Proposal, independent verification, governance decision, and promotion are distinct immutable records. Nothing is promoted automatically. |
+| Protocols | CLI and MCP derive their read surfaces from the same registry, and their overlapping operations are fixture-equivalent. A2A projects Task state; AG-UI projects ordered Run events. Protocol IDs correlate with, but never replace, domain IDs. |
+
+The canonical fixture and operation inventory is [registry.json](registry.json). Closed machine-readable contracts start at [schema/agent-native.schema.json](schema/agent-native.schema.json).
+
+## CLI
+
+Run the CLI from the repository root:
+
+```sh
+npm run sg -- discover --format json
+npm run sg -- resolve sg:profile/editorial-reference-profile --format json
+npm run sg -- claims sg:profile/editorial-reference-profile --format json
+npm run sg -- context sg:profile/editorial-reference-profile --format json
+npm run sg -- ops --format json
+```
+
+The executable can also be invoked directly:
+
+```sh
+./scripts/sg.mjs discover --format json
+```
+
+All machine-mode results are deterministic JSON on stdout. A failed request returns `{ "ok": false }`, a stable failure code, and a nonzero exit status.
+
+### Commands
+
+| Command | Use |
+| --- | --- |
+| `discover` | Read the self-description, immutable manifest identity, conformance receipt, protocol surfaces, and operation summaries. |
+| `resolve <StableRef-or-VersionID>` | Resolve one immutable fixture record. Supplying a VersionID requires an exact version match. |
+| `claims <StableRef>` | Inspect related Claim, Evidence, Validation, Governance, and Policy records without collapsing them into one status. |
+| `context <StableRef>` | Build a deterministic, bounded context package with a snapshot, member manifest, policy, and cache key. |
+| `ops` | List all operation specifications, including internal governed mutations and their exposure metadata. |
+
+## MCP
+
+Start the stdio MCP server with:
+
+```sh
+npm run sg:mcp
+```
+
+An MCP client can launch it directly:
+
+```json
+{
+  "mcpServers": {
+    "stylegallery": {
+      "command": "node",
+      "args": ["/absolute/path/to/StyleGallery/scripts/sg-mcp.mjs"]
+    }
+  }
+}
+```
+
+The MCP surface is read-only. It derives exposed tools from the same registry used by the CLI and accepts only operations marked `read_only: true` with `effect_class: "NONE"`.
+
+Read-only tools:
+
+- `discover`
+- `resolve`
+- `claims`
+- `context`
+- `ops`
+- `retrieve`
+
+Resources:
+
+- `sg://self` - self-description and conformance profile
+- `sg://manifest` - immutable fixture manifest
+- `sg://object/{encoded StableRef}` - one record resolved through the common registry
+
+Registered mutation operations such as `proposal.create`, `proposal.verify`, `proposal.decide`, `proposal.promote`, `task.create`, `run.start`, `effect.record`, and `effect.reconcile` are deliberately absent from MCP.
+
+## Internal Governed Operations
+
+The eight mutation operations are pure internal reducers. They validate and return new immutable records; they do not persist files, alter the fixture registry, or make external calls automatically.
+
+Code that intentionally owns governance and persistence can invoke the common registry:
+
+```js
+import { invokeRegistryOperation } from "../../scripts/agent-native/registry.mjs";
+
+const result = invokeRegistryOperation("task.create", {
+  stable_ref: "sg:task/example",
+  intent: { operation: "resolve" },
+  required_result: { type: "profile_projection" }
+});
+```
+
+Capabilities deny by default when no operation scope is granted. An external effect without a durable connector receipt remains `UNCERTAIN`; only receipt-backed reconciliation may mark it `COMMITTED`.
+
+## Conformance And Trust Boundary
+
+`discover` reports `fixture_verified` only when an immutable execution receipt:
+
+- binds the executable fixture StableRef and VersionID;
+- records the exact command and canonical result digest;
+- carries a passing validator identity and version; and
+- has a VersionID that matches its own canonical content.
+
+This proves fixture-level executable conformance. It does not claim production authorization, remote transport security, exactly-once external effects, or independent organizational review.
+
+## A2A And AG-UI
+
+- A2A exposes an Agent Card and projects StyleGallery Task states to A2A v1 task states.
+- AG-UI projects a Run into ordered `RUN`, text, tool, state, and terminal events.
+- Protocol bindings keep A2A/AG-UI identifiers separate from StyleGallery Task and Run identifiers.
+
+These are projections, not additional sources of business logic. The operation registry and execution records remain canonical.
+
+## Verification
+
+Run the complete agent-native contract suite:
+
+```sh
+npm run test:agent-native
+```
+
+The suite covers identity and manifest tampering, epistemic separation, capability denial, receipt-derived effect state, causal DAG cycles, governed-learning forgery, CLI behavior, real MCP SDK transport, CLI-to-MCP equivalence, schemas, A2A, AG-UI, retrieval, and conformance receipts.
+
+When root routing or this governance entry changes, also run:
+
+```sh
+node scripts/validate-links.mjs --json
+node scripts/validate-ia.mjs --json
+node scripts/validate-governance.mjs --json
+```
+
+## Navigation
+
+Parent: [Consumer Reference](../index.md).
+Next: [Repository governance](../../GOVERNANCE.md).