Add message intent classification (edit/info/help) before routing

Introduces a two-LLM-call pipeline: the first call classifies the user's
message intent as "edit", "info", or "help". Edit messages proceed through
the existing routing → edit → propose flow. Info messages get a generated
response about site content from the manifest. Help messages get a
templated capabilities overview. This handles open-ended questions like
"What can I do?" or "What does my site have on it?" which previously
had no path through the system.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

shared/src/schemas/index.ts

@@ -142,6 +142,13 @@ export const editJobPayloadSchema = z.discriminatedUnion('kind', [
 ]);
 export type EditJobPayload = z.infer<typeof editJobPayloadSchema>;
 
+// ── Classification output (first LLM call) ──
+export const classificationSchema = z.object({
+  intent: z.enum(['edit', 'info', 'help']),
+  reason: z.string(),
+});
+export type ClassificationOutput = z.infer<typeof classificationSchema>;
+
 // ── Routing output (LLM structured output) ──
 export const routingOutputSchema = z.object({
   repo_relative_path: z.string(),