[Capability] Harden structuredContent extraction (#417)

* Address review feedback on resource_link

Three follow-ups from review of the resource_link work:

* PromptResultFormatter: add a regression test proving the optional
  fields of a typed resource_link block (title, description, mimeType,
  size, annotations, _meta) survive formatting. The delegation to
  ResourceLink::fromArray() already landed while rebasing onto the
  PromptResultFormatter refactor, but the existing test only supplied
  uri/name and so could not catch a reintroduced field drop.

* ResourceLink::fromArray(): validate optional fields consistently, the
  way the equivalent ResourceDefinition::fromArray() already does.
  description, mimeType and size now raise InvalidArgumentException
  instead of surfacing a TypeError (or silently coercing, for size's
  (int) cast); annotations goes through Annotations::tryFromArray(),
  which carries the is_array() guard that icons already had; icons uses
  Icon::listFromArray() so a non-array entry is reported in context.

* ToolReference::extractStructuredContent(): never emit a list as
  structuredContent. The Content guard added earlier fixed only one
  instance of the real invariant - structuredContent must be a JSON
  object, and a PHP list can never be one. Tool::fromArray() already
  enforces the matching rule by rejecting an outputSchema whose type is
  not "object", so this aligns the runtime path with it. The test
  asserting the opposite for an array-typed outputSchema contradicted
  the phpstan type and the fromArray() check introduced alongside it in
  the same commit, so it is replaced by tests for the two list shapes.

* Apply the structuredContent object rule to object results too

`extractStructuredContent()` guards raw array results against being emitted
as a JSON array, but the object branch handed back whatever `json_decode()`
produced. A `JsonSerializable` returning a list or a scalar slipped straight
through, producing exactly the `structuredContent` the array guard exists to
prevent — and a return value that contradicts the method's own
`array<string, mixed>|null` signature.

Check the decoded value before returning it, and cover the object branch,
which had no tests at all.

* Document structured tool output

`outputSchema` and `structuredContent` were undocumented: the tool return
value docs covered only the `content` side, and the schema generation
section is about tool parameters.

Add a "Structured Output" subsection covering how to declare the schema,
which return values populate `structuredContent`, and why a list has to be
wrapped in a key to get structured output at all.

* Gate structuredContent on the negotiated protocol revision

SEP-2106 (revision 2026-07-28) widens structuredContent to any JSON value;
earlier revisions require an object. Resolve the revision per request and
apply the matching rule, and warn when a declared outputSchema yields none.

* Resolve the protocol revision on RequestContext

Keeps the `_meta`-then-session lookup in one place instead of per handler,
and exposes the negotiated revision to tool handlers.

* Document reading the negotiated revision from RequestContext

* Warn when a self-built CallToolResult carries an invalid structuredContent

Returning a CallToolResult opts out of the extraction rules, so the value is
still sent unchanged — but a JSON array is not valid before SEP-2106.
This commit is contained in:
Christopher Hertel 2026-08-15 04:03:20 +02:00 • committed by GitHub
commit becce29c41
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -10,6 +10,7 @@ All notable changes to `mcp/sdk` will be documented in this file.
* Add `annotations` support to `ImageContent` (constructor, `fromArray()`, `fromFile()`, `fromString()`, `jsonSerialize()`), matching `TextContent` and `AudioContent`.
* Add client-side `roots/list` handler (`ListRootsRequestHandler` + `RootsCallbackInterface`) and `Client::sendRootsListChanged()`, plus server-side `ClientGateway::listRoots()` / `supportsRoots()` and `ListRootsResult::fromArray()`.
* Add `ClientGateway::supportsSampling()`, so a tool can check the client's advertised capabilities before issuing a `sampling/createMessage` request instead of asking and catching the refusal. Matches the existing `supportsRoots()` and `supportsElicitation()`.
* [BC Break] Gate `structuredContent` on the negotiated protocol revision: `ToolReference::extractStructuredContent()` takes an optional `ProtocolVersion` and, for revisions predating SEP-2106 (`2025-11-25` and earlier, where `structuredContent` must be a JSON object), returns `null` for a tool result that is a PHP list or an object serializing to a JSON array. From `2026-07-28` on both are emitted as-is. Objects serializing to a scalar and arrays holding `Content` instances are never emitted, in any revision. `CallToolHandler` resolves the revision from the request's `_meta` (modern era) or the session (handshake era) and falls back to the strictest rule; it logs a warning when a tool declares an `outputSchema` but returns a value that cannot be sent, and when a self-built `CallToolResult` carries a `structuredContent` the revision does not allow (that one is passed through unchanged). Tools returning a list against an older client keep their JSON-encoded value in `content`; they just no longer advertise an invalid `structuredContent`.
* Add `Mcp\Schema\Content\ResourceLink` for the spec's `resource_link` content block (protocol revision 2025-06-18+), letting tool results and prompt messages reference a resource by URI/name without embedding its contents. Accepted anywhere `resource` (`EmbeddedResource`) content is (de)serialized: `CallToolResult::fromArray()`, `PromptMessage::fromArray()`, and `PromptResultFormatter`.
* Negotiate the protocol revision during the `initialize` handshake: the server echoes a revision it supports and counter-offers `ProtocolVersion::latestHandshake()` otherwise (`Builder::setProtocolVersion()` pins it to exactly one), and the client fails the handshake on a counter-offer it cannot speak rather than continuing on an unagreed revision. Adds `Client::getProtocolVersion()`, the `2026-07-28` revision, and the era helpers on `ProtocolVersion` — revisions from `2026-07-28` on have no `initialize`, so they are excluded from negotiation and from `ProtocolVersionMiddleware`'s default supported set.
* Add sampling with tools support: sampling requests now accept tools and tool-choice preferences, messages support tool-use/tool-result content blocks and multiple content blocks, and clients can advertise the `sampling.context` and `sampling.tools` capabilities.