fix(apps): address review of app-rendered elicitations

Eight findings from the Copilot review, all accepted.

Ownership across connections was the substantive one, in three parts. A
mid-session transport close reaches teardown only through
`clearAndAnnouncePendingPeerRequests`, whose emptiness check looked at the
native queues alone — so a lone app elicitation survived a dropped connection
with its modal still open; the check now counts the active set too. The web
bridge factory resolves its client at call time, so an entry queued by a
replaced `InspectorClient` could rebuild against the *next* one and answer
through a different server; the host now drops every entry synchronously on
the swap (`AppElicitationController.failAll`) rather than racing core's
awaited teardown. And the request id was only monotonic within one client, so
each replacement client's first request was `app-elicitation-1` — settling
that id could resolve the wrong server's request. It now carries a
per-instance prefix.

Result validation now parses the whole value with `ElicitResultSchema` before
the requested-schema check, instead of inspecting `action` by hand: a
`{ action: "decline", content: { x: {} } }` used to pass, though the standard
result permits only primitives and string arrays in `content`.

The dialog's accessible name now includes the request id and the prompt — two
concurrent elicitations can be for the same app URI, and the previous name
made those indistinguishable, which is what the comment beside it claimed to
prevent.

`observeAppCapabilities` takes a minimal structural `onmessage` contract
(generic over the message type, since a handler typed for a narrower message
is not assignable to one typed for `unknown`), removing the double cast at the
call site.

`smoke:web:elicit` now fails on an uncaught page error after a successful
drive, and captures the async half of that class off the console channel, as
its two sibling smokes do — it could otherwise print OK over a broken bundle.

Adds the tests each fix needs, including the two the review asked for
directly: that `advertiseElicitation` reaches the `AppBridge` constructor, and
that it is absent by default.

Signed-off-by: cliffhall <cliff@futurescale.com>
This commit is contained in:
cliffhall 2026-08-22 19:43:44 -04:00
commit 7c77c2b15f
13 changed files with 286 additions and 38 deletions

View file

@ -850,6 +850,20 @@ function App() {
},
[appElicitationController],
);
// Request ownership must not cross connections. `elicitationBridgeFactory`
// resolves its client at call time, so an entry queued by a previous
// InspectorClient would otherwise be able to rebuild against the NEXT one and
// read (or answer) through a different server. Core aborts its own pending
// app requests during teardown, but that teardown is awaited asynchronously
// while the swap here is synchronous — so drop them on the swap instead of
// racing it. The one client alive at a time owns every live entry.
useEffect(() => {
return () => {
appElicitationController.failAll(
new Error("Connection replaced before the app answered"),
);
};
}, [appElicitationController, inspectorClient]);
const [managedToolsState, setManagedToolsState] =
useState<ManagedToolsState | null>(null);

View file

@ -226,11 +226,12 @@ function AppElicitationFrame({
{/* `aria-label` and the `data-*` attributes go on the CONTENT (the
role="dialog" element), not the Root — the Root is only a portal
wrapper, so a name placed there never reaches the dialog. Named per
request rather than per prompt: two concurrent elicitations can carry
the same message, and two identically-named dialogs are what
`landmark-unique` rejects. */}
request rather than per prompt or per app: two concurrent
elicitations can be for the SAME app URI with the SAME message, and
only the request id tells those two dialogs apart — for
`landmark-unique` and for anyone navigating by screen reader. */}
<Modal.Content
aria-label={`Elicitation rendered by ${entry.resourceUri}`}
aria-label={`Elicitation ${entry.requestId} — ${entry.params.message} — rendered by ${entry.resourceUri}`}
data-app-elicitation-status={status}
data-testid="app-elicitation"
>

View file

@ -70,6 +70,30 @@ describe("appCapabilities (#1854)", () => {
expect(appAdvertisesElicitation(bridge)).toBe(false);
});
it("ignores frames that are not shaped like ui/initialize at all", () => {
// The frame comes from sandboxed view code, so nothing about its shape is
// guaranteed — a non-object, or an initialize with no params, must not
// throw on the way through to the bridge's own handler.
const bridge = makeBridge({});
const { transport, inner } = makeTransport();
observeAppCapabilities(bridge, transport);
transport.onmessage?.("not-a-frame" as unknown as JSONRPCMessage);
transport.onmessage?.(null as unknown as JSONRPCMessage);
transport.onmessage?.({
jsonrpc: "2.0",
id: 1,
method: "ui/initialize",
} as unknown as JSONRPCMessage);
transport.onmessage?.({
jsonrpc: "2.0",
id: 2,
method: "ui/initialize",
params: null,
} as unknown as JSONRPCMessage);
expect(appAdvertisesElicitation(bridge)).toBe(false);
expect(inner).toHaveBeenCalledTimes(4);
});
it("keeps bridges independent", () => {
const a = makeBridge({});
const b = makeBridge({});

View file

@ -1,5 +1,4 @@
import type { AppBridge } from "@modelcontextprotocol/ext-apps/app-bridge";
import type { Transport } from "@modelcontextprotocol/client";
/**
* The app capabilities exactly as the view sent them, per bridge.
@ -22,10 +21,41 @@ import type { Transport } from "@modelcontextprotocol/client";
*/
const rawAppCapabilities = new WeakMap<AppBridge, Record<string, unknown>>();
/** Shape of the `ui/initialize` frame this reads — nothing else is touched. */
interface UiInitializeFrame {
method?: unknown;
params?: { appCapabilities?: unknown };
/**
* The only part of a transport this needs: the inbound-message callback.
*
* Structural, and generic over the message type, on purpose. ext-apps'
* `PostMessageTransport` implements the SDK *v1* `Transport` — a different
* nominal type from the v2 client's, though runtime-identical — so naming
* either would force a cast at the call site. Generic rather than
* `(message: unknown)` because a handler typed for a narrower message is not
* assignable to one typed for `unknown` (contravariance), which would put the
* cast back; `rest` is `never[]` for the same reason, accepting any trailing
* parameter list without claiming to know it. The message is only ever read
* through {@link recordAdvertisedCapabilities}, which narrows from `unknown`.
*/
export interface MessageObservable<TMessage = unknown> {
onmessage?: (message: TMessage, ...rest: never[]) => void;
}
/**
* Record `params.appCapabilities` if `message` is a `ui/initialize` frame.
* Narrows from `unknown` rather than asserting a shape: the frame is whatever
* the sandboxed view posted.
*/
function recordAdvertisedCapabilities(
bridge: AppBridge,
message: unknown,
): void {
if (typeof message !== "object" || message === null) return;
const frame = message as { method?: unknown; params?: unknown };
if (frame.method !== "ui/initialize") return;
if (typeof frame.params !== "object" || frame.params === null) return;
const advertised = (frame.params as { appCapabilities?: unknown })
.appCapabilities;
if (typeof advertised === "object" && advertised !== null) {
rawAppCapabilities.set(bridge, advertised as Record<string, unknown>);
}
}
/**
@ -36,20 +66,14 @@ interface UiInitializeFrame {
* wraps. That ordering is safe: the view cannot send `ui/initialize` until the
* host has pushed its HTML into the sandbox, which happens later still.
*/
export function observeAppCapabilities(
export function observeAppCapabilities<TMessage>(
bridge: AppBridge,
transport: Transport,
transport: MessageObservable<TMessage>,
): void {
const inner = transport.onmessage?.bind(transport);
transport.onmessage = (message, extra) => {
const frame = message as UiInitializeFrame;
if (frame.method === "ui/initialize") {
const advertised = frame.params?.appCapabilities;
if (typeof advertised === "object" && advertised !== null) {
rawAppCapabilities.set(bridge, advertised as Record<string, unknown>);
}
}
inner?.(message, extra);
transport.onmessage = (message: TMessage, ...rest: never[]) => {
recordAdvertisedCapabilities(bridge, message);
inner?.(message, ...rest);
};
}

View file

@ -158,6 +158,33 @@ describe("createAppBridgeFactory", () => {
}
});
it("does not advertise hostCapabilities.elicitation by default (#1854)", async () => {
// An App-tool frame is never handed an elicitation, so claiming the host
// capability there would tell those apps something untrue about the host.
const factory = createAppBridgeFactory({
getClient: () => fakeClient,
readResource: vi.fn(),
});
await factory(makeIframe(), { kind: "tool", tool });
expect(bridgeInstances[0].ctorArgs[2]).not.toHaveProperty("elicitation");
});
it("advertises hostCapabilities.elicitation when opted in (#1854)", async () => {
// The value reaches the AppBridge CONSTRUCTOR, which is what the bridge
// echoes in its `ui/initialize` response — an app reads it there to decide
// whether this host will forward an elicitation to it at all.
const factory = createAppBridgeFactory({
advertiseElicitation: true,
getClient: () => fakeClient,
readResource: vi.fn(),
});
await factory(makeIframe(), {
kind: "resource",
resourceUri: "ui://demo/pick.html",
});
expect(bridgeInstances[0].ctorArgs[2]).toMatchObject({ elicitation: {} });
});
it("on sandboxready, reads the UI resource, wraps the html with the per-app CSP, and echoes the approved sandbox config", async () => {
const readResource = vi.fn().mockResolvedValue(
uiResource("<h1>weather</h1>", {

View file

@ -8,7 +8,7 @@ import type {
McpUiHostCapabilities,
McpUiResourceMeta,
} from "@modelcontextprotocol/ext-apps/app-bridge";
import type { Client, Transport } from "@modelcontextprotocol/client";
import type { Client } from "@modelcontextprotocol/client";
import type {
EmbeddedResource,
Implementation,
@ -373,7 +373,7 @@ export function createAppBridgeFactory(
// Record the view's raw `ui/initialize` capabilities before the bridge's
// own schema strips the keys it predates (#1854). Must follow `connect`,
// which is what installs the handler this wraps.
observeAppCapabilities(bridge, transport as unknown as Transport);
observeAppCapabilities(bridge, transport);
return bridge;
};
}

View file

@ -108,6 +108,33 @@ describe("AppElicitationController (#1854)", () => {
expect(controller.getEntries()).toHaveLength(0);
});
it("failAll drops every entry so a renderer cannot outlive its connection", async () => {
// The host calls this when the InspectorClient is replaced: the bridge
// factory resolves its client at call time, so an entry left queued could
// otherwise rebuild against the NEXT connection and answer through a
// different server.
const controller = new AppElicitationController();
const listener = vi.fn();
const first = request(controller, "a");
const second = request(controller, "b");
controller.subscribe(listener);
controller.failAll(new Error("connection replaced"));
await expect(first).rejects.toThrow(/connection replaced/);
await expect(second).rejects.toThrow(/connection replaced/);
expect(controller.getEntries()).toHaveLength(0);
expect(listener).toHaveBeenCalledTimes(1);
});
it("failAll on an empty queue notifies nobody", () => {
const controller = new AppElicitationController();
const listener = vi.fn();
controller.subscribe(listener);
controller.failAll(new Error("nothing to drop"));
expect(listener).not.toHaveBeenCalled();
});
it("stops notifying an unsubscribed listener", () => {
const controller = new AppElicitationController();
const listener = vi.fn();

View file

@ -96,4 +96,20 @@ export class AppElicitationController {
fail(requestId: string, error: Error): void {
this.take(requestId)?.reject(error);
}
/**
* Drop every queued request at once.
*
* For the host's "these entries no longer belong to the live connection"
* moment (a replaced `InspectorClient`): a renderer that outlived its client
* would rebuild against the *next* one and could read or answer through a
* different server.
*/
failAll(error: Error): void {
const dropped = this.entries;
if (dropped.length === 0) return;
this.entries = [];
this.emit();
for (const entry of dropped) entry.reject(error);
}
}

View file

@ -215,7 +215,7 @@ describe("appElicitation negotiation (#1854)", () => {
params,
null as unknown as ElicitResult,
),
).toMatch(/non-object/);
).toMatch(/invalid elicitation result/);
});
it("rejects an unknown action", () => {
@ -223,7 +223,19 @@ describe("appElicitation negotiation (#1854)", () => {
validateAppElicitResult(provider, params, {
action: "maybe",
} as unknown as ElicitResult),
).toMatch(/unknown elicitation action/);
).toMatch(/invalid elicitation result/);
});
it("rejects a decline whose content the standard result forbids", () => {
// Checking `action` alone would wave this through: `ElicitResult` permits
// only primitives and string arrays in `content`, so a nested object
// would reach the server as a result it can legitimately reject.
expect(
validateAppElicitResult(provider, params, {
action: "decline",
content: { x: {} },
} as unknown as ElicitResult),
).toMatch(/invalid elicitation result/);
});
it("rejects an accept with no usable content", () => {
@ -237,7 +249,7 @@ describe("appElicitation negotiation (#1854)", () => {
action: "accept",
content: [] as unknown as Record<string, never>,
} as ElicitResult),
).toMatch(/without a content object/);
).toMatch(/invalid elicitation result|without a content object/);
});
it("rejects content that does not match the requested schema", () => {

View file

@ -405,6 +405,38 @@ describe("app-rendered elicitation routing (#1854)", () => {
});
});
describe("request-scoped ids", () => {
it("mints ids that are unique across client instances, not just within one", async () => {
// A host may hold ONE renderer across replacement clients (the web client
// rebuilds its InspectorClient on a settings change). A per-client
// counter alone would give each new client's first request the same id,
// and settling it would resolve the wrong server's request.
const ids: string[] = [];
const renderer: AppElicitationRenderer = async (request) => {
ids.push(request.requestId);
return { action: "cancel" };
};
const first = new ElicitTransport();
const clientA = await connectClient({
transport: first,
appElicitation: renderer,
});
await first.elicit(40, appParams());
await clientA.disconnect();
const second = new ElicitTransport();
const clientB = await connectClient({
transport: second,
appElicitation: renderer,
});
await second.elicit(41, appParams());
await clientB.disconnect();
expect(ids).toHaveLength(2);
expect(ids[0]).not.toBe(ids[1]);
});
});
describe("teardown", () => {
it("aborts a pending app elicitation on disconnect", async () => {
const transport = new ElicitTransport();
@ -427,5 +459,31 @@ describe("app-rendered elicitation routing (#1854)", () => {
// abandoned it, and the connection it belonged to is gone.
expect(client.getPendingElicitations()).toHaveLength(0);
});
it("aborts a pending app elicitation on a mid-session transport close", async () => {
// The `onclose` route reaches teardown only through
// `clearAndAnnouncePendingPeerRequests`, whose emptiness check used to
// look at the native queues alone — so a lone app elicitation survived a
// dropped connection and left its modal on screen.
const transport = new ElicitTransport();
let aborted = false;
const client = await connectClient({
transport,
appElicitation: (request) =>
new Promise<ElicitResult>((_resolve, reject) => {
request.signal.addEventListener("abort", () => {
aborted = true;
reject(new Error("aborted"));
});
}),
});
void transport.elicit(31, appParams()).catch(() => {});
await vi.waitFor(() => expect(aborted).toBe(false));
transport.onclose?.();
await vi.waitFor(() => expect(aborted).toBe(true));
expect(client.getPendingElicitations()).toHaveLength(0);
});
});
});

View file

@ -6,6 +6,7 @@ import type {
jsonSchemaValidator,
ServerCapabilities,
} from "@modelcontextprotocol/client";
import { ElicitResultSchema } from "@modelcontextprotocol/core";
import { MCP_APP_MIME_TYPE, UI_EXTENSION_KEY } from "./extensions.js";
/**
@ -149,8 +150,13 @@ export function isFormElicitation(params: ElicitRequest["params"]): boolean {
return mode === undefined || mode === "form";
}
/** The three actions a completed elicitation may report. */
const ELICIT_ACTIONS = ["accept", "decline", "cancel"] as const;
/**
* The result contract a completed elicitation must satisfy — the standard MCP
* one, not a hand-rolled subset. Checking `action` by hand would accept, for
* instance, `{ action: "decline", content: { x: {} } }`, whose content the
* schema forbids, and hand the server a result it can reject.
*/
const ELICIT_RESULT_ERROR = "App returned an invalid elicitation result";
/**
* Validates what an app returned before it is handed back to the server.
@ -170,16 +176,13 @@ export function validateAppElicitResult(
params: ElicitRequest["params"],
result: ElicitResult,
): string | undefined {
if (typeof result !== "object" || result === null) {
return "App returned a non-object elicitation result";
}
const action = (result as { action?: unknown }).action;
if (!ELICIT_ACTIONS.includes(action as (typeof ELICIT_ACTIONS)[number])) {
return `App returned an unknown elicitation action: ${JSON.stringify(action)}`;
const parsed = ElicitResultSchema.safeParse(result);
if (!parsed.success) {
return `${ELICIT_RESULT_ERROR}: ${parsed.error.issues[0]?.message ?? "does not match ElicitResult"}`;
}
// decline / cancel carry no content — they are complete as they stand.
if (action !== "accept") return undefined;
const content = (result as { content?: unknown }).content;
if (parsed.data.action !== "accept") return undefined;
const content = parsed.data.content;
if (
typeof content !== "object" ||
content === null ||

View file

@ -584,6 +584,17 @@ export class InspectorClient extends InspectorClientEventTarget {
* elicitations cannot resolve through each other's bridges.
*/
private appElicitationSeq = 0;
/**
* Per-instance prefix for {@link appElicitationSeq}.
*
* A counter alone is not enough: a host may hold ONE renderer across
* replacement clients (the web client rebuilds its `InspectorClient` on a
* settings change), and every fresh instance would otherwise mint
* `app-elicitation-1` for its first request. Settling that id would then
* resolve — or discard — whichever connection's request the host happened to
* be keying.
*/
private readonly appElicitationPrefix = `app-elicitation-${globalThis.crypto?.randomUUID?.() ?? Math.random().toString(36).slice(2)}`;
/**
* Abort controllers for app-rendered elicitations still awaiting an answer.
* Aborted alongside the native pending queue on disconnect, so a rendered app
@ -1755,7 +1766,12 @@ export class InspectorClient extends InspectorClientEventTarget {
private clearAndAnnouncePendingPeerRequests(): void {
if (
this.pendingSamples.length === 0 &&
this.pendingElicitations.length === 0
this.pendingElicitations.length === 0 &&
// App-rendered elicitations (#1854) are not in either array — they live in
// the host's renderer — so an emptiness check that ignores them lets a
// mid-session close (the `onclose` route, which reaches teardown only
// through here) leave a modal open for a connection that is gone.
this.activeAppElicitations.size === 0
) {
return;
}
@ -2946,7 +2962,7 @@ export class InspectorClient extends InspectorClientEventTarget {
this.activeAppElicitations.add(controller);
try {
const result = await renderer({
requestId: `app-elicitation-${++this.appElicitationSeq}`,
requestId: `${this.appElicitationPrefix}-${++this.appElicitationSeq}`,
resourceUri,
params: request.params,
signal: controller.signal,

View file

@ -60,6 +60,9 @@ const PORT = process.env.SMOKE_WEB_ELICIT_PORT ?? "6296";
const TOKEN = "smoke-web-elicit-token";
const TOOL = "app_choose_option";
const SHOT_DIR = process.env.SMOKE_SCREENSHOT_DIR;
// The async half of the uncaught-crash class. Kept identical to
// smoke-web-browser.mjs / smoke-web-app.mjs.
const FATAL_CONSOLE = /^Uncaught\b|Failed to fetch dynamically imported module/;
const servers = [];
let browser = null;
@ -235,10 +238,20 @@ try {
viewport: { width: 1280, height: 900 },
});
// Uncaught *synchronous* page errors, and their *async* twin (an unhandled
// rejection or a failed dynamic import), which Chromium reports on the
// console channel instead. Both are hard failures; every other console error
// is only a diagnostic, so benign font/React noise cannot flake CI. Same
// split as smoke:web:browser and smoke:web:app, which document it at length.
const pageErrors = [];
const consoleErrors = [];
page.on("pageerror", (err) =>
pageErrors.push(err instanceof Error ? err.message : String(err)),
);
page.on("console", (msg) => {
if (msg.type() === "error") consoleErrors.push(msg.text());
});
const fatalConsole = () => consoleErrors.filter((m) => FATAL_CONSOLE.test(m));
const drive = async () => {
// ── 1. Negotiated: the server's app answers the elicitation ────────────
@ -304,13 +317,26 @@ try {
try {
await Promise.race([web.whenChildExits(), drive()]);
} catch (err) {
const diagnostics = [
...pageErrors,
...fatalConsole().map((m) => `console: ${m}`),
];
await fail(
`${err instanceof Error ? err.message : String(err)}${
pageErrors.length ? ` — page errors: ${pageErrors.join("; ")}` : ""
diagnostics.length
? ` — page diagnostics: ${diagnostics.join("; ")}`
: ""
}`,
);
}
// A drive that reached all of its assertions still fails if the page threw on
// the way: without this the smoke prints OK over a broken bundle.
const fatal = [...pageErrors, ...fatalConsole()];
if (fatal.length > 0) {
await fail(`page logged uncaught error(s): ${fatal.join("; ")}`);
}
console.log("smoke:web:elicit OK — app-rendered and native paths both drive");
await shutdown();
} catch (err) {