diff --git a/server/auth-realtime/CLAUDE.md b/server/auth-realtime/CLAUDE.md
index 38fc35bc..56c5e868 100644
--- a/server/auth-realtime/CLAUDE.md
+++ b/server/auth-realtime/CLAUDE.md
@@ -202,10 +202,27 @@ stale.
   [src/test-setup.ts](src/test-setup.ts) turns that warning into a failure so
   it cannot be missed: vitest prints intercepted stderr only under the
   verbose reporter, so in a piped or CI run the warning is invisible.
-- `src/schema/` is out of scope for the suite: it is a declarative mirror of
-  web's tiptap extension list, and the behaviour that depends on it is
-  covered through the transformer in `ydocument` and `operations`. It is
-  excluded from coverage.
+- `src/schema/` is the one place that departs from 1:1 pairing: the whole
+  directory is tested through [src/schema/index.test.ts](src/schema/index.test.ts)
+  rather than a test file per node. The files are one unit — a single
+  editor schema, re-exported through `index.ts` — and the tests that
+  matter run against the assembled whole: a node round-tripped through
+  the transformer proves it is registered, which no per-file test can.
+  The per-node HTML assertions are data tables in the same file for the
+  same reason.
+
+  What the suite pins there is the contract with the two services on
+  either side. **Round trips** (prosemirror JSON → Y.Doc → prosemirror
+  JSON) prove the schema knows each node, because a node it does not know
+  is silently dropped on the way through — including the non-string
+  attributes (`queries`, `thresholds`) that `cloneXmlElement` exists to
+  protect. **The uid list** on the `UniqueID` extension is checked to
+  contain every addressable block and to name only extensions that
+  exist, because `operations.ts` addresses blocks by uid and a block
+  missing from that list is one the assistant can never edit. **The
+  `data-type` strings** are pinned because this service never serializes
+  to HTML — those callbacks exist purely to stay identical to web's
+  definitions, so nothing else would notice them drifting apart.
 
 ### Naming
 
diff --git a/server/auth-realtime/src/schema/index.test.ts b/server/auth-realtime/src/schema/index.test.ts
new file mode 100644
index 00000000..ddf75a66
--- /dev/null
+++ b/server/auth-realtime/src/schema/index.test.ts
@@ -0,0 +1,932 @@
+import { describe, it } from "vitest"
+import {
+	CalloutBlock,
+	CodeBlock,
+	CodeBlockTitle,
+	getEditorExtensions,
+	ParameterList,
+	ParameterListHeader,
+	ParameterListItem,
+	ParameterListItemHeader,
+	ParameterListItemHeaderTitle,
+	ParameterListItemHeaderType,
+	SplitDocumentation,
+	SplitDocumentationLeftSide,
+	SplitDocumentationRightSide,
+	TitledCodeBlock,
+	AddedMark,
+	CommentMark,
+	DeletedMark,
+} from "./index.js"
+import { FigmaBlock } from "./figma.js"
+import { ImageBlock } from "./image.js"
+import { MermaidBlock } from "./mermaid.js"
+import { MetricBlock, MetricGrid } from "./metric.js"
+import { transformer } from "../ydocument.js"
+
+interface PMNode {
+	type: string
+	attrs?: Record<string, unknown>
+	content?: PMNode[]
+	text?: string
+	marks?: { type: string; attrs?: Record<string, unknown> }[]
+}
+
+// the schema's real job is to survive the trip every document takes
+// between the editor and core: prosemirror JSON into a Y.Doc and back
+// out. A node the schema does not know is silently dropped on the way
+// through, so a round trip is what proves it is registered.
+function roundTrip(block: PMNode): PMNode {
+	const ydoc = transformer.toYdoc(
+		{ type: "doc", content: [block] },
+		"content",
+	)
+	const back = transformer.fromYdoc(ydoc, "content") as PMNode
+
+	const first = back.content?.[0]
+	if (!first) {
+		throw new Error("the round trip produced an empty document")
+	}
+
+	return first
+}
+
+function paragraph(uid: string, text: string): PMNode {
+	return {
+		type: "paragraph",
+		attrs: { uid },
+		content: [{ type: "text", text }],
+	}
+}
+
+function heading(uid: string, text: string): PMNode {
+	return {
+		type: "heading",
+		attrs: { uid, level: 2 },
+		content: [{ type: "text", text }],
+	}
+}
+
+function metricBlock(uid: string): PMNode {
+	return {
+		type: "metricBlock",
+		attrs: {
+			uid,
+			title: "Requests",
+			queries: [{ expr: "up", legend: "up" }],
+			thresholds: [{ value: 10, color: "red" }],
+			decimals: 2,
+			unitType: "short",
+		},
+	}
+}
+
+function titledCodeBlock(uid: string): PMNode {
+	return {
+		type: "titledCodeBlock",
+		attrs: { uid },
+		content: [
+			{
+				type: "codeBlockTitle",
+				attrs: { uid: `${uid}-title` },
+				content: [{ type: "text", text: "restart.sh" }],
+			},
+			{
+				type: "codeBlock",
+				attrs: { uid: `${uid}-code`, language: "bash" },
+				content: [
+					{
+						type: "text",
+						text: "systemctl restart",
+					},
+				],
+			},
+		],
+	}
+}
+
+describe("getEditorExtensions", () => {
+	it("gives every extension a unique name", ({ expect }) => {
+		const names = getEditorExtensions().map((e) => e.name)
+
+		expect(new Set(names).size).toBe(names.length)
+	})
+
+	// a document cannot be parsed at all without them, and they are the
+	// two the custom blocks never declare themselves
+	it("includes the document and text nodes every schema needs", ({
+		expect,
+	}) => {
+		const names = getEditorExtensions().map((e) => e.name)
+
+		expect(names).toContain("doc")
+		expect(names).toContain("text")
+	})
+
+	// operations.ts addresses every block by its uid, so a block missing
+	// from this list is one the assistant can never edit
+	it.for([
+		{ name: "callout", input: "calloutBlock" },
+		{ name: "metric", input: "metricBlock" },
+		{ name: "metric grid", input: "metricGrid" },
+		{ name: "image", input: "imageBlock" },
+		{ name: "figma", input: "figmaBlock" },
+		{ name: "mermaid", input: "mermaidBlock" },
+		{ name: "titled code", input: "titledCodeBlock" },
+		{ name: "split documentation", input: "splitDocumentation" },
+		{
+			name: "parameter list",
+			input: "splitDocumentationParameterList",
+		},
+	])("gives the $name block a uid", ({ input }, { expect }) => {
+		const uid = getEditorExtensions().find(
+			(e) => e.name === "uniqueID",
+		)
+		const types = (uid?.options as { types?: string[] } | undefined)
+			?.types
+
+		expect(types).toContain(input)
+	})
+
+	// a name here that no extension answers to would silently leave that
+	// block without uids
+	it("lists only registered extensions as uid carriers", ({ expect }) => {
+		const extensions = getEditorExtensions()
+		const names = new Set(extensions.map((e) => e.name))
+		const uid = extensions.find((e) => e.name === "uniqueID")
+		const types =
+			(uid?.options as { types?: string[] } | undefined)
+				?.types ?? []
+
+		expect(types.filter((t) => !names.has(t))).toEqual([])
+	})
+})
+
+describe("editor schema", () => {
+	describe("round trip", () => {
+		it.for([
+			{
+				name: "a paragraph",
+				input: paragraph("p1", "Restart it"),
+			},
+			{ name: "a heading", input: heading("h1", "Runbook") },
+			{
+				name: "an image",
+				input: {
+					type: "imageBlock",
+					attrs: {
+						uid: "i1",
+						src: "https://example/x.png",
+						alt: "x",
+						width: 400,
+					},
+				},
+			},
+			{
+				name: "a figma embed",
+				input: {
+					type: "figmaBlock",
+					attrs: {
+						uid: "f1",
+						src: "https://figma/x",
+						width: 800,
+						height: 600,
+					},
+				},
+			},
+			{
+				name: "a horizontal rule",
+				input: {
+					type: "horizontalRule",
+					attrs: { uid: "r1" },
+				},
+			},
+		])("keeps $name intact", ({ input }, { expect }) => {
+			expect(roundTrip(input)).toEqual(input)
+		})
+
+		// the reason cloneXmlElement exists: Y.XmlElement.clone() drops
+		// everything that is not a string, which would strip a metric
+		// block's queries on every fork and merge
+		it("keeps a metric block's array and number attributes", ({
+			expect,
+		}) => {
+			const result = roundTrip(metricBlock("m1"))
+
+			expect(result.attrs?.queries).toEqual([
+				{ expr: "up", legend: "up" },
+			])
+			expect(result.attrs?.thresholds).toEqual([
+				{ value: 10, color: "red" },
+			])
+			expect(result.attrs?.decimals).toBe(2)
+		})
+
+		it("keeps a mermaid block's source text", ({ expect }) => {
+			const result = roundTrip({
+				type: "mermaidBlock",
+				attrs: { uid: "d1" },
+				content: [
+					{
+						type: "text",
+						text: "graph TD;\n  A-->B;",
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.text).toBe(
+				"graph TD;\n  A-->B;",
+			)
+		})
+
+		it("keeps a callout's icon and its nested paragraph", ({
+			expect,
+		}) => {
+			const result = roundTrip({
+				type: "calloutBlock",
+				attrs: { uid: "c1", icon: "lucide:siren" },
+				content: [
+					paragraph("c1-p", "Page the on-call"),
+				],
+			})
+
+			expect(result.attrs?.icon).toBe("lucide:siren")
+			expect(result.content?.[0]?.content?.[0]?.text).toBe(
+				"Page the on-call",
+			)
+		})
+
+		it("keeps the blocks inside a metric grid", ({ expect }) => {
+			const result = roundTrip({
+				type: "metricGrid",
+				attrs: { uid: "g1" },
+				content: [metricBlock("m1"), metricBlock("m2")],
+			})
+
+			expect(result.content).toHaveLength(2)
+			expect(result.content?.[1]?.attrs?.uid).toBe("m2")
+		})
+
+		it("keeps a titled code block's title and language", ({
+			expect,
+		}) => {
+			const result = roundTrip(titledCodeBlock("t1"))
+
+			expect(result.content?.[0]?.content?.[0]?.text).toBe(
+				"restart.sh",
+			)
+			expect(result.content?.[1]?.attrs?.language).toBe(
+				"bash",
+			)
+		})
+
+		// the widest node in the schema: a heading and prose on the
+		// left, code or metrics on the right
+		it("keeps both sides of a split documentation block", ({
+			expect,
+		}) => {
+			const result = roundTrip({
+				type: "splitDocumentation",
+				attrs: { uid: "s1" },
+				content: [
+					{
+						type: "splitDocumentationLeftSide",
+						attrs: { uid: "s1-l" },
+						content: [
+							heading(
+								"s1-h",
+								"Endpoint",
+							),
+							paragraph(
+								"s1-p",
+								"Returns a page",
+							),
+						],
+					},
+					{
+						type: "splitDocumentationRightSide",
+						attrs: { uid: "s1-r" },
+						content: [
+							titledCodeBlock("s1-c"),
+						],
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.type).toBe(
+				"splitDocumentationLeftSide",
+			)
+			expect(result.content?.[1]?.type).toBe(
+				"splitDocumentationRightSide",
+			)
+			expect(result.content?.[0]?.content?.[0]?.type).toBe(
+				"heading",
+			)
+		})
+
+		it("keeps a parameter list's header and items", ({
+			expect,
+		}) => {
+			const result = roundTrip({
+				type: "splitDocumentationParameterList",
+				attrs: { uid: "pl1" },
+				content: [
+					{
+						type: "splitDocumentationParameterListHeader",
+						attrs: { uid: "pl1-h" },
+						content: [
+							{
+								type: "text",
+								text: "Parameters",
+							},
+						],
+					},
+					{
+						type: "splitDocumentationParameterListItem",
+						attrs: { uid: "pl1-i" },
+						content: [
+							{
+								type: "splitDocumentationParameterListItemHeader",
+								attrs: {
+									uid: "pl1-ih",
+								},
+								content: [
+									{
+										type: "splitDocumentationParameterListItemHeaderTitle",
+										attrs: {
+											uid: "pl1-t",
+										},
+										content: [
+											{
+												type: "text",
+												text: "limit",
+											},
+										],
+									},
+									{
+										type: "splitDocumentationParameterListItemHeaderType",
+										attrs: {
+											uid: "pl1-ty",
+										},
+										content: [
+											{
+												type: "text",
+												text: "number",
+											},
+										],
+									},
+								],
+							},
+							paragraph(
+								"pl1-d",
+								"How many to return",
+							),
+						],
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.content?.[0]?.text).toBe(
+				"Parameters",
+			)
+			expect(
+				result.content?.[1]?.content?.[0]?.content?.[0]
+					?.content?.[0]?.text,
+			).toBe("limit")
+		})
+	})
+
+	describe("marks", () => {
+		// comment marks carry the id the comment sidebar resolves; a
+		// dropped attribute orphans the thread
+		it("keeps a comment mark's id", ({ expect }) => {
+			const result = roundTrip({
+				type: "paragraph",
+				attrs: { uid: "p1" },
+				content: [
+					{
+						type: "text",
+						text: "flagged",
+						marks: [
+							{
+								type: "comment",
+								attrs: {
+									commentId: "cmt-1",
+								},
+							},
+						],
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.marks?.[0]?.type).toBe(
+				"comment",
+			)
+			expect(
+				result.content?.[0]?.marks?.[0]?.attrs
+					?.commentId,
+			).toBe("cmt-1")
+		})
+
+		// the diff view marks additions and deletions inline
+		it.for([
+			{ name: "added", input: "added" },
+			{ name: "deleted", input: "deleted" },
+		])("keeps a $name mark", ({ input }, { expect }) => {
+			const result = roundTrip({
+				type: "paragraph",
+				attrs: { uid: "p1" },
+				content: [
+					{
+						type: "text",
+						text: "changed",
+						marks: [{ type: input }],
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.marks?.[0]?.type).toBe(
+				input,
+			)
+		})
+
+		it.for([
+			{ name: "bold", input: "bold" },
+			{ name: "italic", input: "italic" },
+			{ name: "strike", input: "strike" },
+			{ name: "underline", input: "underline" },
+			{ name: "code", input: "code" },
+		])("keeps a $name mark", ({ input }, { expect }) => {
+			const result = roundTrip({
+				type: "paragraph",
+				attrs: { uid: "p1" },
+				content: [
+					{
+						type: "text",
+						text: "styled",
+						marks: [{ type: input }],
+					},
+				],
+			})
+
+			expect(result.content?.[0]?.marks?.[0]?.type).toBe(
+				input,
+			)
+		})
+
+		it("keeps a link's href", ({ expect }) => {
+			const result = roundTrip({
+				type: "paragraph",
+				attrs: { uid: "p1" },
+				content: [
+					{
+						type: "text",
+						text: "docs",
+						marks: [
+							{
+								type: "link",
+								attrs: {
+									href: "https://example/docs",
+								},
+							},
+						],
+					},
+				],
+			})
+
+			expect(
+				result.content?.[0]?.marks?.[0]?.attrs?.href,
+			).toBe("https://example/docs")
+		})
+	})
+})
+
+// the HTML side of the schema. This service never serializes a document
+// to HTML — the transformer works on prosemirror JSON — so these
+// callbacks exist only to keep the definitions identical to web's. A
+// data-type that drifts apart is a block web writes and cannot parse
+// back, which is why the strings are pinned here rather than left to the
+// two copies to agree by luck.
+type Attrs = Record<string, unknown>
+
+interface AttributeSpec {
+	default?: unknown
+	parseHTML?: (element: {
+		getAttribute(name: string): string | null
+	}) => unknown
+	renderHTML?: (attrs: Attrs) => Attrs
+}
+
+interface NodeSpec {
+	renderHTML?: (
+		this: { options: Attrs },
+		props: { HTMLAttributes: Attrs; node?: { attrs: Attrs } },
+	) => [string, Attrs, ...unknown[]]
+	parseHTML?: (this: { options: Attrs }) => { tag: string }[]
+	addAttributes?: (this: {
+		options: Attrs
+	}) => Record<string, AttributeSpec>
+}
+
+function specOf(node: { config: unknown }): NodeSpec {
+	return node.config as NodeSpec
+}
+
+function render(
+	node: { config: unknown },
+	attrs: Attrs = {},
+	options: Attrs = {},
+): [string, Attrs, ...unknown[]] {
+	const rendered = specOf(node).renderHTML?.call(
+		{ options },
+		{ HTMLAttributes: {}, node: { attrs } },
+	)
+	if (!rendered) {
+		throw new Error("the node declares no renderHTML")
+	}
+
+	return rendered
+}
+
+function parseSelector(node: { config: unknown }): string | undefined {
+	return specOf(node).parseHTML?.call({ options: {} })[0]?.tag
+}
+
+function attributesOf(node: {
+	config: unknown
+}): Record<string, AttributeSpec> {
+	return specOf(node).addAttributes?.call({ options: {} }) ?? {}
+}
+
+describe("schema html contract", () => {
+	it.for([
+		{
+			name: "callout",
+			input: CalloutBlock,
+			expected: "callout-block",
+		},
+		{
+			name: "metric grid",
+			input: MetricGrid,
+			expected: "metric-grid",
+		},
+		{
+			name: "metric",
+			input: MetricBlock,
+			expected: "metric-block",
+		},
+		{ name: "figma", input: FigmaBlock, expected: "figma-block" },
+		{
+			name: "mermaid",
+			input: MermaidBlock,
+			expected: "mermaid-block",
+		},
+		{ name: "code", input: CodeBlock, expected: "code-block" },
+		{
+			name: "code title",
+			input: CodeBlockTitle,
+			expected: "code-block-title",
+		},
+		{
+			name: "titled code",
+			input: TitledCodeBlock,
+			expected: "titled-code-block",
+		},
+		{
+			name: "split documentation",
+			input: SplitDocumentation,
+			expected: "split-documentation",
+		},
+		{
+			name: "split left side",
+			input: SplitDocumentationLeftSide,
+			expected: "split-documentation-left-side",
+		},
+		{
+			name: "split right side",
+			input: SplitDocumentationRightSide,
+			expected: "split-documentation-right-side",
+		},
+		{
+			name: "parameter list",
+			input: ParameterList,
+			expected: "split-documentation-parameter-list",
+		},
+		{
+			name: "parameter list header",
+			input: ParameterListHeader,
+			expected: "split-documentation-parameter-list-header",
+		},
+		{
+			name: "parameter list item",
+			input: ParameterListItem,
+			expected: "split-documentation-parameter-list-item",
+		},
+		{
+			name: "parameter list item header",
+			input: ParameterListItemHeader,
+			expected: "split-documentation-parameter-list-item-header",
+		},
+		{
+			name: "parameter list item title",
+			input: ParameterListItemHeaderTitle,
+			expected: "split-documentation-parameter-list-item-header-title",
+		},
+		{
+			name: "parameter list item type",
+			input: ParameterListItemHeaderType,
+			expected: "split-documentation-parameter-list-item-header-type",
+		},
+	])(
+		"tags the $name block as $expected and parses it back",
+		({ input, expected }, { expect }) => {
+			expect(render(input)[1]["data-type"]).toBe(expected)
+			expect(parseSelector(input)).toContain(expected)
+		},
+	)
+
+	it.for([
+		{ name: "image", input: ImageBlock, expected: "img" },
+		{ name: "mermaid", input: MermaidBlock, expected: "pre" },
+		{ name: "code", input: CodeBlock, expected: "pre" },
+		{ name: "callout", input: CalloutBlock, expected: "div" },
+	])(
+		"renders the $name block as a $expected element",
+		({ input, expected }, { expect }) => {
+			expect(render(input)[0]).toBe(expected)
+		},
+	)
+
+	describe("callout icon", () => {
+		it("reads the icon off data-icon", ({ expect }) => {
+			const icon = attributesOf(CalloutBlock).icon
+
+			expect(
+				icon?.parseHTML?.({
+					getAttribute: (name) =>
+						name === "data-icon"
+							? "lucide:zap"
+							: null,
+				}),
+			).toBe("lucide:zap")
+		})
+
+		it("writes the icon back to data-icon", ({ expect }) => {
+			const icon = attributesOf(CalloutBlock).icon
+
+			expect(
+				icon?.renderHTML?.({ icon: "lucide:siren" }),
+			).toEqual({
+				"data-icon": "lucide:siren",
+			})
+		})
+
+		it("defaults to the plain text icon", ({ expect }) => {
+			expect(attributesOf(CalloutBlock).icon?.default).toBe(
+				"lucide:text",
+			)
+		})
+	})
+
+	describe("figma attributes", () => {
+		// the width and height arrive as strings on the element and are
+		// stored as numbers, so a document round-trips as numbers
+		it.for([
+			{ name: "width", input: "width" },
+			{ name: "height", input: "height" },
+		])("parses $name into a number", ({ input }, { expect }) => {
+			const attr = attributesOf(FigmaBlock)[input]
+
+			expect(
+				attr?.parseHTML?.({
+					getAttribute: () => "800",
+				}),
+			).toBe(800)
+		})
+
+		it.for([
+			{ name: "src", input: "src" },
+			{ name: "width", input: "width" },
+			{ name: "height", input: "height" },
+		])(
+			"parses a missing $name as null",
+			({ input }, { expect }) => {
+				const attr = attributesOf(FigmaBlock)[input]
+
+				expect(
+					attr?.parseHTML?.({
+						getAttribute: () => null,
+					}),
+				).toBeNull()
+			},
+		)
+
+		it.for([
+			{ name: "src", input: "src", expected: "data-src" },
+			{
+				name: "width",
+				input: "width",
+				expected: "data-width",
+			},
+			{
+				name: "height",
+				input: "height",
+				expected: "data-height",
+			},
+		])(
+			"writes $name back to $expected",
+			({ input, expected }, { expect }) => {
+				const attr = attributesOf(FigmaBlock)[input]
+
+				expect(
+					attr?.renderHTML?.({ [input]: 800 }),
+				).toEqual({
+					[expected]: 800,
+				})
+			},
+		)
+
+		// an unset value writes no attribute at all rather than an
+		// empty one, which is what keeps the markup clean
+		it.for([
+			{ name: "src", input: "src" },
+			{ name: "width", input: "width" },
+			{ name: "height", input: "height" },
+		])(
+			"writes nothing when $name is unset",
+			({ input }, { expect }) => {
+				const attr = attributesOf(FigmaBlock)[input]
+
+				expect(attr?.renderHTML?.({})).toEqual({})
+			},
+		)
+	})
+
+	describe("marks", () => {
+		it.for([
+			{
+				name: "added",
+				input: AddedMark,
+				expected: "data-added",
+			},
+			{
+				name: "deleted",
+				input: DeletedMark,
+				expected: "data-deleted",
+			},
+		])(
+			"flags a $name range with $expected",
+			({ input, expected }, { expect }) => {
+				expect(render(input)[1][expected]).toBe("true")
+				expect(parseSelector(input)).toContain(expected)
+			},
+		)
+
+		it("writes a comment's id to data-comment-id", ({ expect }) => {
+			const attr = attributesOf(CommentMark).commentId
+
+			expect(
+				attr?.renderHTML?.({ commentId: "cmt-1" }),
+			).toEqual({
+				"data-comment-id": "cmt-1",
+			})
+		})
+
+		it("reads a comment's id off data-comment-id", ({ expect }) => {
+			const attr = attributesOf(CommentMark).commentId
+
+			expect(
+				attr?.parseHTML?.({
+					getAttribute: (name) =>
+						name === "data-comment-id"
+							? "cmt-1"
+							: null,
+				}),
+			).toBe("cmt-1")
+		})
+	})
+
+	describe("code block language", () => {
+		it("puts the language on the inner code element", ({
+			expect,
+		}) => {
+			const rendered = render(
+				CodeBlock,
+				{ language: "bash" },
+				{ languageClassPrefix: "language-" },
+			)
+			const code = rendered[2] as [
+				string,
+				Attrs,
+				...unknown[],
+			]
+
+			expect(code[1].class).toBe("language-bash")
+		})
+
+		it("leaves the class unset when no language is chosen", ({
+			expect,
+		}) => {
+			const rendered = render(
+				CodeBlock,
+				{},
+				{ languageClassPrefix: "language-" },
+			)
+			const code = rendered[2] as [
+				string,
+				Attrs,
+				...unknown[],
+			]
+
+			expect(code[1].class).toBeNull()
+		})
+	})
+})
+
+describe("split documentation orientation", () => {
+	it.for([
+		{ name: "set", input: "true", expected: true },
+		{ name: "absent", input: null, expected: false },
+		{ name: "some other value", input: "yes", expected: false },
+	])(
+		"reads an orientation that is $name as $expected",
+		({ input, expected }, { expect }) => {
+			const attr = attributesOf(SplitDocumentation).inversed
+
+			expect(
+				attr?.parseHTML?.({
+					getAttribute: () => input,
+				}),
+			).toBe(expected)
+		},
+	)
+
+	it("writes the orientation back only when it is inverted", ({
+		expect,
+	}) => {
+		const attr = attributesOf(SplitDocumentation).inversed
+
+		expect(attr?.renderHTML?.({ inversed: true })).toEqual({
+			"data-inversed": "true",
+		})
+		expect(attr?.renderHTML?.({ inversed: false })).toEqual({})
+	})
+
+	it("defaults to the uninverted orientation", ({ expect }) => {
+		expect(attributesOf(SplitDocumentation).inversed?.default).toBe(
+			false,
+		)
+	})
+})
+
+describe("comment mark rendering", () => {
+	it("renders a comment as a span carrying its attributes", ({
+		expect,
+	}) => {
+		const rendered = specOf(CommentMark).renderHTML?.call(
+			{ options: {} },
+			{ HTMLAttributes: { "data-comment-id": "cmt-1" } },
+		)
+
+		expect(rendered?.[0]).toBe("span")
+		expect(rendered?.[1]["data-comment-id"]).toBe("cmt-1")
+	})
+})
+
+describe("code block without configured options", () => {
+	// the prefix comes from the extension's options, and a caller that
+	// renders the node without them still has to produce a usable class
+	it("falls back to an unprefixed language class", ({ expect }) => {
+		const rendered = render(CodeBlock, { language: "bash" }, {})
+		const code = rendered[2] as [string, Attrs, ...unknown[]]
+
+		expect(code[1].class).toBe("bash")
+	})
+})
+
+describe("code block options", () => {
+	// the definition extends CodeBlockLowlight, so its own defaults are
+	// layered over whatever the parent extension supplied — and it has to
+	// hold up when there is no parent to layer over
+	it.for([
+		{ name: "a parent extension", input: true },
+		{ name: "no parent at all", input: false },
+	])("resolves its defaults with $name", ({ input }, { expect }) => {
+		const addOptions = (
+			CodeBlock.config as {
+				addOptions?: (this: {
+					parent?: () => Record<string, unknown>
+				}) => Record<string, unknown>
+			}
+		).addOptions
+
+		const options = addOptions?.call(
+			input
+				? { parent: () => ({ tabSize: 8 }) }
+				: { parent: undefined },
+		)
+
+		expect(options?.languageClassPrefix).toBe("language-")
+		expect(options?.tabSize).toBe(input ? 8 : 2)
+	})
+})
diff --git a/server/auth-realtime/src/schema/index.ts b/server/auth-realtime/src/schema/index.ts
index c0672293..e4286dc8 100644
--- a/server/auth-realtime/src/schema/index.ts
+++ b/server/auth-realtime/src/schema/index.ts
@@ -115,8 +115,11 @@ export function getEditorExtensions(): Extensions {
 		UniqueID.configure({
 			types: contentExtensionsWithIDs.map((v) => v.name),
 			attributeName: "uid",
-			// annotated because tiptap declares generateID as
-			// returning any, which propagates to the arrow
+			// NOCOV: UniqueID calls this only while an editor is
+			// applying a transaction, which never happens in this
+			// service — it transforms documents, it does not edit
+			// them. Annotated because tiptap declares generateID as
+			// returning any, which propagates to the arrow.
 			generateID: (): string => nanoid(),
 		}),
 	]
diff --git a/server/auth-realtime/vitest.config.ts b/server/auth-realtime/vitest.config.ts
index d3cd81ba..99f2b6f2 100644
--- a/server/auth-realtime/vitest.config.ts
+++ b/server/auth-realtime/vitest.config.ts
@@ -36,11 +36,6 @@ export default defineConfig({
 				// loaded through node's --import before the app
 				// graph exists, so it can hold no imports of ours
 				"src/sentry.ts",
-				// the editor schema mirrors web's tiptap
-				// extension list; it is configuration, and the
-				// behaviour that depends on it is covered through
-				// the transformer in ydocument and operations
-				"src/schema/**",
 				"src/**/test-helpers.ts",
 				"src/test-setup.ts",
 			],
@@ -49,10 +44,10 @@ export default defineConfig({
 			// genuinely untestable branches stay visible in the
 			// report, marked with NOCOV comments in place
 			thresholds: {
-				statements: 98.4,
-				branches: 96.3,
-				functions: 98.4,
-				lines: 98.4,
+				statements: 98.5,
+				branches: 96.6,
+				functions: 98.5,
+				lines: 98.5,
 			},
 		},
 	},