[skills] Document structured outputs and file attachments

- Structured outputs: json_schema response_format across formats,
  legacy json form, streaming accumulation, per-format pages
- File attachments: content-part arrays, image_url and file parts,
  discovery through modalities.input and vision/file-input tags
- Retrieval intents and prompt signals for both

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: jerilynzheng

generated/build-from-skills.manifest.json

@@ -1,6 +1,6 @@
 {
   "version": 1,
-  "generatedAt": "2026-08-30T22:26:59.912Z",
+  "generatedAt": "2026-08-30T23:33:07.371Z",
   "templates": [
     {
       "template": "agents/ai-architect.md.tmpl",

generated/skill-catalog.md

@@ -1,7 +1,7 @@
 # Skill Catalog
 
 > Auto-generated by `scripts/generate-catalog.ts` — do not edit manually.
-> Generated: 2026-08-30T22:27:03.861Z
+> Generated: 2026-08-30T23:33:09.994Z
 > Skills: 35
 
 ## Table of Contents

generated/skill-manifest.json

@@ -1,5 +1,5 @@
 {
-  "generatedAt": "2026-08-30T22:27:02.744Z",
+  "generatedAt": "2026-08-30T23:33:09.108Z",
   "version": 2,
   "skills": {
     "access-protected-vercel-deployment": {
@@ -231,7 +231,8 @@
           "credit balance",
           "safety identifier",
           "reasoning effort",
-          "tool calling"
+          "tool calling",
+          "structured outputs"
         ],
         "noneOf": [
           "cloudflare ai gateway",
@@ -258,7 +259,9 @@
           "find a model by modality, capability, price, or data retention",
           "check AI Gateway credit balance or generation cost",
           "configure reasoning or extended thinking across providers and API formats",
-          "add tool calling or function calling across API formats"
+          "add tool calling or function calling across API formats",
+          "get structured JSON output matching a schema",
+          "send images or PDFs to a model"
         ],
         "entities": [
           "AI Gateway",

skills/ai-gateway/SKILL.md

@@ -43,6 +43,7 @@ metadata:
       - "safety identifier"
       - "reasoning effort"
       - "tool calling"
+      - "structured outputs"
     noneOf:
       - "cloudflare ai gateway"
       - "aws api gateway"
@@ -94,6 +95,8 @@ retrieval:
     - check AI Gateway credit balance or generation cost
     - configure reasoning or extended thinking across providers and API formats
     - add tool calling or function calling across API formats
+    - get structured JSON output matching a schema
+    - send images or PDFs to a model
   entities:
     - AI Gateway
     - AI Gateway Credits

skills/ai-gateway/references/routing.md

@@ -137,6 +137,29 @@ Web search is also available as a built-in tool for supported models instead of
 
 Read the per-format tool-calling page before writing schemas; argument serialization and streaming deltas differ by format.
 
+## Structured outputs across API formats
+
+Each language API format accepts a JSON schema for the response, applied to whichever provider serves the request:
+
+- Chat Completions: `response_format` with `type: 'json_schema'` and a named `json_schema.schema`. A legacy `type: 'json'` form exists for backward compatibility; use `json_schema` for new code.
+- Responses, OpenResponses, and Anthropic Messages: each has its own structured-outputs page with the format's parameter shape.
+- AI SDK: use its structured-data generation APIs; load the `ai-sdk` skill for exact exports.
+
+With streaming, structured output arrives as ordinary content deltas; accumulate the full stream and parse once it finishes. Model support varies, so verify a schema-heavy flow with a live request before building on it.
+
+Docs: <https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/structured-outputs>
+
+## File attachments
+
+Models with image or file input accept attachments as a content-part array in place of a plain string message, mixed with text parts in one message:
+
+- Images: an `image_url` part carrying a data URI, with an optional `detail` field.
+- PDFs: a `file` part carrying the document as a data URI.
+
+Discovery: the model's `modalities.input` array and the `vision` and `file-input` tags in `/v1/models`. Each API format has its own attachments page, and part types differ by format, so read the one matching the request shape before writing payloads.
+
+Docs: <https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-chat-completions/images>
+
 ## Automatic prompt caching
 
 ```ts