Reach tasks attached to topics, and read Etherpad pads

Testing against a local instance turned up four things the server was
getting wrong, all of them invisible against the live account because the
data that exposes them had never been produced there.

`GET /lessons/{id}/tasks` returns a bare array, not the `{data,total}`
envelope every sibling endpoint uses, so `.data` was undefined and a
topic's tasks silently vanished. Its items also carry no id at all —
`LessonLinkedTaskResponse` has no id property — which leaves a
topic-attached task unidentifiable: it is not a task element on the
course page, and once past due it is in neither task list. So its
submission, and its grade, could not be reached by any route. That is 18
of 60 tasks on the real account, now reachable: the ids come off the
legacy topic page, where each task is linked as `/homework/{id}`.

The types said `id: string` and `status: TaskStatus` on something that
has neither, which is what let this stay quiet; `LessonLinkedTask` and
`ResolvedTask` now say what is actually there.

Collaborative text editor elements come back with `content: {}`, and the
tool said their contents were unavailable. They are available: the
content-element endpoint returns the pad url *and* an Etherpad session
cookie, and the pad exports itself as text to whoever holds it. No API
key needed. Pads are now shown by get_board and indexed for search.

The store's file digest covered id and size on the grounds that file
records are immutable. `PATCH /file/rename/{id}` renames one in place,
so a rename was reported as nothing at all.

Finally, get_board reported an unpublished board as "no permission",
which sends the reader hunting for an access problem that is not there.

smoke gains checks for topic tasks and for pads, and no longer assumes a
populated index or a search term that happens to match. 39/39 live-only
and 41/41 index-backed, against both the live instance and a local one.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/core/board.ts

@@ -1,4 +1,6 @@
+import type { Config } from '../config.ts';
 import type { SchulcloudClient } from './client.ts';
+import { fetchPadText } from './etherpad.ts';
 import { SchulcloudApiError } from './client.ts';
 import type { BoardSkeleton, CardResponse, ContentElement, FileRecord } from './types.ts';
 
@@ -22,6 +24,8 @@ export interface AssembledElement {
 	files: FileRecord[];
 	/** Set when this element's files could not be resolved. */
 	fileError?: string;
+	/** What a class actually wrote in a collaborativeTextEditor (Etherpad) pad. */
+	padText?: string;
 	raw: Record<string, unknown>;
 }
 
@@ -52,7 +56,7 @@ export async function assembleBoard(
 	client: SchulcloudClient,
 	boardId: string,
 	schoolId: string,
-	options: { resolveFiles?: boolean } = {},
+	options: { resolveFiles?: boolean; resolvePads?: Config } = {},
 ): Promise<AssembledBoard> {
 	const resolveFiles = options.resolveFiles ?? true;
 
@@ -71,6 +75,13 @@ export async function assembleBoard(
 		await attachFiles(client, assembled, schoolId);
 	}
 
+	// A pad's text is real course content, and nothing else surfaces it: the
+	// board API returns collaborativeTextEditor elements with empty content.
+	// Costs two requests per pad and only when a board has one.
+	if (options.resolvePads) {
+		await attachPadText(options.resolvePads, assembled);
+	}
+
 	const fileCount = assembled
 		.flatMap((column) => column.cards)
 		.flatMap((card) => card.elements)
@@ -121,6 +132,29 @@ function buildElement(element: ContentElement): AssembledElement {
 	return assembled;
 }
 
+/**
+ * Fills in the text of each collaborative text editor element.
+ *
+ * Failures are left silent rather than recorded: unlike a missing attachment,
+ * an unreadable pad is usually an empty one, and the element itself is still
+ * reported.
+ */
+async function attachPadText(config: Config, columns: AssembledColumn[]): Promise<void> {
+	const pads = columns
+		.flatMap((column) => column.cards)
+		.flatMap((card) => card.elements)
+		.filter((element) => element.type === 'collaborativeTextEditor');
+
+	await Promise.all(
+		pads.map(async (element) => {
+			const text = await fetchPadText(config, element.id);
+			if (!text) return;
+			// The element's own title, when it has one, stays as the heading.
+			element.padText = text;
+		}),
+	);
+}
+
 /**
  * Resolves file-bearing elements to file records.
  *

src/core/client.ts

@@ -13,6 +13,7 @@ import type {
 	NewsResponse,
 	Paginated,
 	SubmissionStatus,
+	LessonLinkedTask,
 	TaskContent,
 } from './types.ts';
 
@@ -304,8 +305,22 @@ export class SchulcloudClient {
 		return this.getJson<LessonResponse>(`/api/v3/lessons/${encodeURIComponent(lessonId)}`);
 	}
 
-	getLessonTasks(lessonId: string): Promise<Paginated<TaskContent>> {
-		return this.getJson<Paginated<TaskContent>>(`/api/v3/lessons/${encodeURIComponent(lessonId)}/tasks`);
+	/**
+	 * A lesson's tasks.
+	 *
+	 * Returns a bare array, not the `{data, total}` envelope every other list
+	 * endpoint uses — checked against both the live instance and a local 33.40.
+	 * Typing it as `Paginated` made `.data` undefined, which silently dropped
+	 * every task attached to a topic: they vanished from get_lesson, get_task
+	 * reported them as non-existent, and their submissions — grades included —
+	 * could not be reached at all. The envelope branch is kept in case the
+	 * endpoint is ever normalised to match its siblings.
+	 */
+	async getLessonTasks(lessonId: string): Promise<LessonLinkedTask[]> {
+		const body = await this.getJson<LessonLinkedTask[] | Paginated<LessonLinkedTask>>(
+			`/api/v3/lessons/${encodeURIComponent(lessonId)}/tasks`,
+		);
+		return Array.isArray(body) ? body : (body.data ?? []);
 	}
 
 	// --- column boards ---------------------------------------------------

src/core/crawl.ts

@@ -1,3 +1,4 @@
+import type { Config } from '../config.ts';
 import { assembleBoard, type AssembledBoard } from './board.ts';
 import type { SchulcloudClient } from './client.ts';
 import { htmlToText, normalizeObjectId } from './text.ts';
@@ -90,6 +91,11 @@ export interface CrawlOptions {
 	includeLessonContents?: boolean;
 	/** Resolve board file elements to file records. */
 	includeFiles?: boolean;
+	/**
+	 * Read the text of collaborative text editor (Etherpad) pads, which needs a
+	 * second credentialled hop outside the API. Omit to leave pads unread.
+	 */
+	config?: Config;
 	courseConcurrency?: number;
 	boardConcurrency?: number;
 	onProgress?: (done: number, total: number, label: string) => void;
@@ -203,7 +209,10 @@ async function crawlCourse(
 	await forEachLimited(boardIds, options.boardConcurrency ?? 4, async (boardId) => {
 		let assembled: AssembledBoard;
 		try {
-			assembled = await assembleBoard(client, boardId, options.schoolId, { resolveFiles: includeFiles });
+			assembled = await assembleBoard(client, boardId, options.schoolId, {
+				resolveFiles: includeFiles,
+				resolvePads: options.config,
+			});
 		} catch (error) {
 			// Record rather than swallow: a dropped board used to disappear from the
 			// index while the crawl still reported success, which is how a 20-card
@@ -222,6 +231,9 @@ async function crawlCourse(
 				parts.push(card.title);
 				for (const element of card.elements) {
 					if (element.text) parts.push(htmlToText(element.text));
+					// Pad contents are course material like any other; without this they
+					// are unsearchable, and a pad is often where the actual group work is.
+					if (element.padText) parts.push(element.padText);
 					if (element.url) parts.push(element.url);
 					for (const record of element.files) {
 						files.push({

src/core/etherpad.ts

@@ -0,0 +1,100 @@
+import type { Config } from '../config.ts';
+
+/**
+ * Reads the text of a collaborative text editor (Etherpad) element.
+ *
+ * The board API is no help on its own: a `collaborativeTextEditor` element
+ * comes back with `content: {}` — no pad id, no url, nothing. What a class
+ * actually wrote in it is invisible to every other tool here.
+ *
+ * Two calls recover it, and neither needs Etherpad's own API key (which is a
+ * server-side secret this process has no business holding):
+ *
+ *  1. `GET /api/v3/collaborative-text-editor/content-element/{id}` answers with
+ *     the pad url *and*, in a `Set-Cookie`, an Etherpad `sessionID` — the same
+ *     exchange the web client performs before it embeds the pad.
+ *  2. Etherpad's own `/p/{padId}/export/txt` returns the pad as plain text to
+ *     whoever holds that session.
+ *
+ * The session cookie is only ever sent back to the instance's own host: step 1
+ * returns a url built from the server's `ETHERPAD__PAD_URI`, and a value
+ * pointing anywhere else is refused rather than followed.
+ *
+ * Everything degrades to `undefined`. A pad that cannot be read costs its text,
+ * never the board.
+ */
+export async function fetchPadText(config: Config, elementId: string): Promise<string | undefined> {
+	try {
+		const handle = await fetchPadHandle(config, elementId);
+		if (!handle) return undefined;
+
+		const response = await fetch(`${handle.origin}/etherpad/p/${handle.padId}/export/txt`, {
+			headers: { Cookie: handle.sessionCookie, Accept: 'text/plain' },
+			signal: AbortSignal.timeout(config.requestTimeoutMs),
+		});
+		if (!response.ok) return undefined;
+
+		const text = (await response.text()).trim();
+		// A pad nobody has typed in still exports the placeholder the instance
+		// seeds new pads with; reporting that as content would be a lie.
+		return text.length > 0 && text !== DEFAULT_PAD_TEXT ? text : undefined;
+	} catch {
+		return undefined;
+	}
+}
+
+/** The instance's `DEFAULT_PAD_TEXT`; an untouched pad exports exactly this. */
+const DEFAULT_PAD_TEXT = 'Schreib etwas!';
+
+interface PadHandle {
+	origin: string;
+	padId: string;
+	sessionCookie: string;
+}
+
+async function fetchPadHandle(config: Config, elementId: string): Promise<PadHandle | undefined> {
+	const response = await fetch(
+		`${config.baseUrl}/api/v3/collaborative-text-editor/content-element/${encodeURIComponent(elementId)}`,
+		{
+			headers: { Authorization: `Bearer ${config.jwt}`, Accept: 'application/json' },
+			signal: AbortSignal.timeout(config.requestTimeoutMs),
+		},
+	);
+	if (!response.ok) return undefined;
+
+	const body = (await response.json()) as { url?: unknown };
+	if (typeof body.url !== 'string') return undefined;
+
+	const padId = padIdFromUrl(body.url, config.baseUrl);
+	if (!padId) return undefined;
+
+	// `getSetCookie` keeps the header split correctly; a plain `get` would join
+	// several cookies on the commas that appear inside the session list itself.
+	const sessionCookie = response.headers
+		.getSetCookie()
+		.map((cookie) => /^(sessionID=[^;]*)/.exec(cookie)?.[1])
+		.find((value): value is string => Boolean(value));
+	if (!sessionCookie) return undefined;
+
+	return { origin: new URL(config.baseUrl).origin, padId, sessionCookie };
+}
+
+/**
+ * The pad id out of the url the server hands back.
+ *
+ * Refuses a url on another host: that url is server-configured, and following
+ * it blindly would send an Etherpad session cookie wherever it pointed.
+ * Group pad ids contain `$`, so the segment is kept exactly as encoded.
+ */
+export function padIdFromUrl(url: string, baseUrl: string): string | undefined {
+	let parsed: URL;
+	try {
+		parsed = new URL(url);
+	} catch {
+		return undefined;
+	}
+	if (parsed.host !== new URL(baseUrl).host) return undefined;
+
+	const segment = /\/etherpad\/p\/([^/?#]+)/.exec(parsed.pathname)?.[1];
+	return segment && segment.length > 0 ? segment : undefined;
+}

src/core/lesson-page.ts

@@ -0,0 +1,84 @@
+import type { Config } from '../config.ts';
+import { decodeEntities } from './text.ts';
+import type { LessonLinkedTask } from './types.ts';
+
+/**
+ * Recovers the ids of tasks that hang off a topic ("Thema").
+ *
+ * `GET /api/v3/lessons/{id}/tasks` returns the topic's tasks — with their name,
+ * description and dates, but **no id**: `LessonLinkedTaskResponse` has no id
+ * field at all, by design rather than by omission. The course page is no help
+ * either, because a topic-attached task is not a task element there; the topic
+ * reports only `numberOfPublishedTasks`.
+ *
+ * The consequence is that those tasks were unreachable: absent from both task
+ * lists once past due (open excludes them, finished only holds what the student
+ * ticked off), absent from the course page, and unidentifiable from the topic.
+ * Their submissions, and so their grades, could not be read at all. On the
+ * account this was written for that is 18 of 60 tasks.
+ *
+ * The legacy topic page links each task as `/homework/{id}`, so it carries the
+ * mapping the API withholds. Like the homework-page scrape this authenticates
+ * by `jwt` **cookie**, hangs off an accessibility attribute rather than
+ * presentation markup, and degrades to an empty list — a markup change costs
+ * the ids again, never an error.
+ */
+
+export interface LessonTaskLink {
+	id: string;
+	name: string;
+}
+
+export async function fetchLessonTaskLinks(
+	config: Config,
+	courseId: string,
+	lessonId: string,
+): Promise<LessonTaskLink[]> {
+	const url = `${config.baseUrl}/courses/${encodeURIComponent(courseId)}/topics/${encodeURIComponent(lessonId)}`;
+	try {
+		const response = await fetch(url, {
+			headers: { Cookie: `jwt=${config.jwt}`, Accept: 'text/html' },
+			signal: AbortSignal.timeout(config.requestTimeoutMs),
+			redirect: 'follow',
+		});
+		// Redirected away means the cookie was not accepted; there is nothing to
+		// parse and nothing worth raising.
+		if (!response.ok || !new URL(response.url).hostname.endsWith(new URL(config.baseUrl).hostname)) {
+			return [];
+		}
+		return parseLessonTaskLinks(await response.text());
+	} catch {
+		return [];
+	}
+}
+
+/** Exported for testing: the parsing is pure and deserves fixtures, not a network. */
+export function parseLessonTaskLinks(html: string): LessonTaskLink[] {
+	// `aria-label="Details der Aufgabe: 'name'"` exists for screen readers, which
+	// makes it far steadier than the surrounding layout.
+	const pattern = /<a href="\/homework\/([0-9a-f]{24})"[^>]*aria-label="[^"']*'([^']*)'/g;
+	const found: LessonTaskLink[] = [];
+	const seen = new Set<string>();
+	for (const match of html.matchAll(pattern)) {
+		const id = match[1] ?? '';
+		if (seen.has(id)) continue;
+		seen.add(id);
+		found.push({ id, name: decodeEntities(match[2] ?? '').trim() });
+	}
+	return found;
+}
+
+/**
+ * Pairs the API's id-less task bodies with the ids scraped from the topic page.
+ *
+ * Matching is by name, which is what both sides agree on. A name the page does
+ * not account for yields a task without an id: still worth reporting (it is
+ * visible to the user), just not something the id-taking tools can open.
+ */
+export function withScrapedIds(tasks: LessonLinkedTask[], links: LessonTaskLink[]): LessonLinkedTask[] {
+	const byName = new Map(links.map((link) => [link.name, link.id]));
+	return tasks.map((task) => {
+		const id = task.id ?? byName.get(task.name?.trim() ?? '');
+		return id ? { ...task, id } : task;
+	});
+}

src/core/types.ts

@@ -71,6 +71,39 @@ export interface TaskContent {
 	status: TaskStatus;
 }
 
+/**
+ * A topic's task as `GET /lessons/{id}/tasks` returns it.
+ *
+ * Deliberately not a `TaskContent`: the response carries no `id` and no
+ * `status` — `LessonLinkedTaskResponse` simply has no id property. Treating it
+ * as a TaskContent made `task.id` undefined at runtime while the type claimed
+ * otherwise, which is how topic-attached tasks went missing in silence.
+ * `core/lesson-page.ts` recovers the ids.
+ */
+export interface LessonLinkedTask {
+	/** Absent from the API; filled in from the topic page when it can be. */
+	id?: string;
+	name: string;
+	description?: string;
+	availableDate?: string;
+	dueDate?: string | null;
+	courseId?: string;
+	courseName?: string;
+	lessonName?: string;
+	private?: boolean;
+	submissionIds?: string[];
+	finishedIds?: string[];
+}
+
+/**
+ * A task as the tools render it, from whichever route found it.
+ *
+ * The task lists and course pages carry a `status`; the topic projection does
+ * not, and carries no id until one is scraped. Every `TaskContent` satisfies
+ * this, so list-derived tasks keep their full detail.
+ */
+export type ResolvedTask = LessonLinkedTask & { status?: TaskStatus };
+
 export interface LessonMetaContent {
 	id: string;
 	name: string;

src/indexer/indexer.ts

@@ -113,6 +113,7 @@ export class Indexer {
 				courseIds: scope === 'full' ? undefined : [scope],
 				includeLessonContents: true,
 				includeFiles: true,
+				config: this.config,
 			});
 
 			const crawlId = await this.store.saveSnapshot(snapshot, scope);

src/mcp/tools/content.ts

@@ -1,11 +1,19 @@
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import { z } from 'zod';
 import type { ServerContext } from '../../context.ts';
+import { SchulcloudApiError } from '../../core/client.ts';
 import { formatBytes } from '../../core/extract.ts';
 import { dueLabel, formatDate, heading, htmlToText, joinSections, normalizeObjectId } from '../../core/text.ts';
 import { assembleBoard, type AssembledBoard, type AssembledElement } from '../../core/board.ts';
 import { forEachLimited } from '../../core/crawl.ts';
-import type { CourseBoardResponse, FileRecord, LessonResponse, TaskContent } from '../../core/types.ts';
+import { fetchLessonTaskLinks, withScrapedIds } from '../../core/lesson-page.ts';
+import type {
+	CourseBoardResponse,
+	FileRecord,
+	LessonLinkedTask,
+	LessonResponse,
+	ResolvedTask,
+} from '../../core/types.ts';
 import { failure, text, toToolError } from './result.ts';
 import { describeSubmission } from './submissions.ts';
 
@@ -55,9 +63,24 @@ export function registerContentTools(server: McpServer, context: ServerContext):
 		async ({ boardId, includeFiles }) => {
 			try {
 				const schoolId = await context.schoolId();
-				const board = await assembleBoard(context.client, boardId, schoolId, { resolveFiles: includeFiles });
+				const board = await assembleBoard(context.client, boardId, schoolId, {
+					resolveFiles: includeFiles,
+					resolvePads: context.config,
+				});
 				return text(formatBoard(board, includeFiles));
 			} catch (error) {
+				// An unpublished board 403s, while the course page lists its title
+				// regardless — the course-board projection does not filter drafts.
+				// Reporting that as "no permission" sends the reader looking for an
+				// access problem that does not exist; a draft is the common cause.
+				if (error instanceof SchulcloudApiError && error.status === 403) {
+					return failure(
+						`Board ${boardId} could not be opened (HTTP 403).\n\n` +
+							`The usual reason is that it is still a draft: an unpublished board is listed on ` +
+							`the course page with its title, but stays closed until the teacher publishes it. ` +
+							`Otherwise this account genuinely has no access to it.`,
+					);
+				}
 				return toToolError(error, `read board ${boardId}`);
 			}
 		},
@@ -80,12 +103,18 @@ export function registerContentTools(server: McpServer, context: ServerContext):
 				const schoolId = await context.schoolId();
 				const [lesson, tasks, files] = await Promise.all([
 					context.client.getLesson(lessonId),
-					context.client.getLessonTasks(lessonId).catch(() => undefined),
+					context.client.getLessonTasks(lessonId).catch(() => []),
 					context.client
 						.listFiles({ storageLocationId: schoolId, parentType: 'lessons', parentId: lessonId })
 						.catch(() => undefined),
 				]);
-				return text(formatLesson(lesson, tasks?.data ?? [], files?.data ?? []));
+				// The task bodies carry no id, so get_task cannot be pointed at them
+				// without the topic page. Only paid for when the topic has tasks.
+				const withIds =
+					tasks.length > 0
+						? withScrapedIds(tasks, await fetchLessonTaskLinks(context.config, lesson.courseId, lessonId))
+						: tasks;
+				return text(formatLesson(lesson, withIds, files?.data ?? []));
 			} catch (error) {
 				return toToolError(error, `read lesson ${lessonId}`);
 			}
@@ -142,7 +171,7 @@ export function registerContentTools(server: McpServer, context: ServerContext):
  * to learn which course the task belongs to (unless told), then read the
  * description off that course's page.
  */
-async function findTask(context: ServerContext, taskId: string, courseId?: string): Promise<TaskContent | undefined> {
+async function findTask(context: ServerContext, taskId: string, courseId?: string): Promise<ResolvedTask | undefined> {
 	if (courseId) {
 		const fromCourse = await taskFromCourse(context, courseId, taskId);
 		if (fromCourse) return fromCourse;
@@ -168,7 +197,7 @@ async function findTask(context: ServerContext, taskId: string, courseId?: strin
 	// back to scanning course pages costs ~26 requests and a few seconds, which
 	// is a fair price for the tool working instead of claiming the id is wrong.
 	const courses = await context.client.listAllCourses().catch(() => []);
-	let found: TaskContent | undefined;
+	let found: ResolvedTask | undefined;
 	await forEachLimited(courses, 6, async (course) => {
 		if (found) return;
 		const fromCourse = await taskFromCourse(context, course.id, taskId);
@@ -181,7 +210,7 @@ async function taskFromCourse(
 	context: ServerContext,
 	courseId: string,
 	taskId: string,
-): Promise<TaskContent | undefined> {
+): Promise<ResolvedTask | undefined> {
 	const board = await context.client.getCourseBoard(courseId).catch(() => undefined);
 	if (!board) return undefined;
 	for (const element of board.elements) {
@@ -189,6 +218,21 @@ async function taskFromCourse(
 			return { ...element.content, courseId, courseName: element.content.courseName ?? board.title };
 		}
 	}
+
+	// A task can hang off a topic rather than the course page, and those are not
+	// listed as task elements — only as a count on the topic. Without this the
+	// task is unreachable: not in the lists (a submitted, past-due task is in
+	// neither open nor finished) and not on the course page either.
+	for (const element of board.elements) {
+		if (element.type !== 'lesson' || !element.content.numberOfPublishedTasks) continue;
+		const links = await fetchLessonTaskLinks(context.config, courseId, element.content.id);
+		if (!links.some((link) => link.id === taskId)) continue;
+		const tasks = await context.client.getLessonTasks(element.content.id).catch(() => []);
+		const match = withScrapedIds(tasks, links).find((task) => task.id === taskId);
+		if (match) {
+			return { ...match, courseId, courseName: match.courseName ?? board.title, lessonName: element.content.name };
+		}
+	}
 	return undefined;
 }
 
@@ -252,6 +296,14 @@ function formatBoard(board: AssembledBoard, includeFiles: boolean): string {
 	]);
 }
 
+/** Indents a pad's body so it reads as quoted content, not as board structure. */
+function indent(body: string): string {
+	return body
+		.split('\n')
+		.map((line) => `  > ${line}`.trimEnd())
+		.join('\n');
+}
+
 function formatElement(element: AssembledElement, includeFiles: boolean): string {
 	switch (element.type) {
 		case 'richText': {
@@ -271,8 +323,16 @@ function formatElement(element: AssembledElement, includeFiles: boolean): string
 			if (element.files.length === 0) return `- ${element.type} element \`${element.id}\` — no files${caption}`;
 			return element.files.map((file) => `- ${formatFileLine(file)}${caption}`).join('\n');
 		}
-		case 'collaborativeTextEditor':
-			return `- Collaborative text document \`${element.id}\`${element.text ? ` — ${element.text}` : ''} (contents not available through the API)`;
+		case 'collaborativeTextEditor': {
+			const title = element.text ? ` — ${element.text}` : '';
+			// The board API returns these with empty content; the text comes from
+			// the pad itself (core/etherpad.ts). Absent means empty or unreadable,
+			// which for a pad is usually "nobody has written in it yet".
+			if (!element.padText) {
+				return `- Collaborative text document \`${element.id}\`${title} (empty, or its contents could not be read)`;
+			}
+			return [`- Collaborative text document \`${element.id}\`${title}:`, indent(element.padText)].join('\n');
+		}
 		case 'externalTool':
 			return `- External tool${element.text ? `: ${element.text}` : ''} \`${element.id}\``;
 		case 'videoConference':
@@ -292,7 +352,7 @@ export function formatFileLine(file: FileRecord): string {
 	return `File: **${file.name}** (\`${file.id}\`, ${file.mimeType}, ${formatBytes(file.size)})${blocked}${pending}`;
 }
 
-function formatLesson(lesson: LessonResponse, tasks: TaskContent[], files: FileRecord[]): string {
+function formatLesson(lesson: LessonResponse, tasks: LessonLinkedTask[], files: FileRecord[]): string {
 	const sections = (lesson.contents ?? []).map((entry) => {
 		const title = entry.title?.trim();
 		const component = entry.component ?? 'unknown';
@@ -316,7 +376,12 @@ function formatLesson(lesson: LessonResponse, tasks: TaskContent[], files: FileR
 		tasks.length > 0 &&
 			joinSections([
 				heading(3, `Tasks in this lesson (${tasks.length})`),
-				tasks.map((task) => `- **${task.name}** (\`${task.id}\`) — ${dueLabel(task.dueDate)}`).join('\n'),
+				tasks
+					.map((task) => {
+						const id = task.id ? ` (\`${task.id}\`)` : '';
+						return `- **${task.name}**${id} — ${dueLabel(task.dueDate)}`;
+					})
+					.join('\n'),
 			]),
 	]);
 }
@@ -336,7 +401,7 @@ function formatLessonComponent(component: string, content: Record<string, unknow
 	return '';
 }
 
-function formatTask(task: TaskContent, files: FileRecord[], submission?: string): string {
+function formatTask(task: ResolvedTask, files: FileRecord[], submission?: string): string {
 	const description = htmlToText(task.description);
 	return joinSections([
 		heading(2, task.name),
@@ -346,7 +411,11 @@ function formatTask(task: TaskContent, files: FileRecord[], submission?: string)
 			task.lessonName ? `- Topic: ${task.lessonName}` : undefined,
 			`- Available from: ${formatDate(task.availableDate)}`,
 			`- Due: ${dueLabel(task.dueDate)}`,
-			`- Submitted: ${task.status.submitted}/${task.status.maxSubmissions}${task.status.graded > 0 ? ', graded' : ''}`,
+			// Absent for a task found through a topic: that projection reports no
+			// counts. The submission section below carries the authoritative state.
+			task.status
+				? `- Submitted: ${task.status.submitted}/${task.status.maxSubmissions}${task.status.graded > 0 ? ', graded' : ''}`
+				: undefined,
 		]
 			.filter(Boolean)
 			.join('\n'),

src/mcp/tools/search.ts

@@ -110,6 +110,9 @@ async function liveSearch(
 		courseIds: courseId ? [courseId] : undefined,
 		includeLessonContents: true,
 		includeFiles: scoped,
+		// Same trade as files: worth two extra requests per pad when the caller
+		// named a course, too slow to do across every course they can see.
+		config: scoped ? context.config : undefined,
 	});
 	const hits = searchSnapshot(snapshot, query, limit);
 

src/mcp/tools/submissions.ts

@@ -1,10 +1,11 @@
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import { z } from 'zod';
 import type { ServerContext } from '../../context.ts';
+import { fetchLessonTaskLinks, withScrapedIds } from '../../core/lesson-page.ts';
 import { forEachLimited } from '../../core/crawl.ts';
 import { fetchSubmissionDetail } from '../../core/homework-page.ts';
 import { dueLabel, heading, joinSections } from '../../core/text.ts';
-import type { FileRecord, SubmissionStatus, TaskContent } from '../../core/types.ts';
+import type { FileRecord, ResolvedTask, SubmissionStatus } from '../../core/types.ts';
 import { formatFileLine } from './content.ts';
 import { text, toToolError } from './result.ts';
 
@@ -52,7 +53,7 @@ export function registerSubmissionTools(server: McpServer, context: ServerContex
 				const [me, tasks] = await Promise.all([context.me(), collectTasks(context, scope, courseId, limit)]);
 				if (tasks.length === 0) return text('No tasks found to check for submissions.');
 
-				const rows: { task: TaskContent; status: SubmissionStatus }[] = [];
+				const rows: { task: ResolvedTask; status: SubmissionStatus }[] = [];
 				const unavailable: string[] = [];
 
 				await forEachLimited(tasks, 5, async (task) => {
@@ -125,7 +126,7 @@ export function formatGradeState(
 	return 'marked graded, but neither a percentage nor feedback was found';
 }
 
-function formatRow({ task, status }: { task: TaskContent; status: SubmissionStatus }): string {
+function formatRow({ task, status }: { task: ResolvedTask; status: SubmissionStatus }): string {
 	const state = status.isSubmitted ? 'submitted' : 'not submitted';
 	// The list does not fetch pages, so it cannot know whether feedback exists;
 	// it says only what the API told it.
@@ -144,8 +145,8 @@ async function collectTasks(
 	scope: 'open' | 'finished' | 'all',
 	courseId: string | undefined,
 	limit: number,
-): Promise<TaskContent[]> {
-	const wanted: TaskContent[] = [];
+): Promise<(ResolvedTask & { id: string })[]> {
+	const wanted: ResolvedTask[] = [];
 	if (scope === 'open' || scope === 'all') {
 		wanted.push(...(await context.client.listTasks({ limit }).catch(() => ({ data: [] }))).data);
 	}
@@ -160,10 +161,24 @@ async function collectTasks(
 		for (const element of board?.elements ?? []) {
 			if (element.type === 'task') wanted.push({ ...element.content, courseId });
 		}
+		// Tasks attached to a topic are not task elements on the course page, so
+		// they have to be asked for per topic. They are the ones most likely to
+		// carry a grade: a task old enough to have been marked is usually old
+		// enough to have dropped out of both task lists.
+		for (const element of board?.elements ?? []) {
+			if (element.type !== 'lesson' || !element.content.numberOfPublishedTasks) continue;
+			const [tasks, links] = await Promise.all([
+				context.client.getLessonTasks(element.content.id).catch(() => []),
+				fetchLessonTaskLinks(context.config, courseId, element.content.id),
+			]);
+			// Only the ones whose id could be recovered: a submission lookup needs it.
+			for (const task of withScrapedIds(tasks, links)) if (task.id) wanted.push({ ...task, courseId });
+		}
 	}
 
 	const seen = new Set<string>();
 	return wanted
+		.filter((task): task is ResolvedTask & { id: string } => Boolean(task.id))
 		.filter((task) => (courseId ? task.courseId === courseId : true))
 		.filter((task) => (seen.has(task.id) ? false : (seen.add(task.id), true)))
 		.slice(0, limit);

src/store/store.ts

@@ -544,9 +544,12 @@ function fileNode(file: CrawledFile): StoredNode {
 			securityCheckStatus: file.record.securityCheckStatus,
 			at: file.at,
 		},
-		// File records are immutable, so identity alone decides change; the size
-		// is included only to catch an upstream record being rewritten in place.
-		digest: digestOf([file.record.id, file.record.size]),
+		// File records are less immutable than they look: `PATCH /file/rename/{id}`
+		// changes the name in place, keeping the id and the size, and teachers do
+		// rename files. Leaving the name out made that invisible to what_changed —
+		// the file simply reappeared under its new name with nothing reported.
+		// Size still catches a record rewritten in place under the same name.
+		digest: digestOf([file.record.id, file.record.name, file.record.size]),
 	};
 }