feat: enhance selectLlmsAndOptions with intelligent look & feel analysis

Incorporates PR 219's intelligent decision logic into PR 229's simpler architecture.
The system now analyzes user prompts to determine if specific look & feel requirements
are described, and only includes instructional text and demo data for plain CRUD apps
without specified styling. This provides contextually appropriate guidance without the
complex slot/catalog plumbing from PR 219.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>

Author: jchris

app/prompts.ts

@@ -91,7 +91,7 @@ export async function selectLlmsAndOptions(
     {
       role: 'system',
       content:
-        'You select which library modules from a catalog should be included AND whether to include instructional UI text and a demo-data button. Read the JSON payload and return JSON with properties: "selected" (array of catalog "name" strings), "instructionalText" (boolean), and "demoData" (boolean). Only choose modules from the catalog. Include any libraries already used in history. Respond with JSON only.',
+        'You select which library modules from a catalog should be included AND whether to include instructional UI text and a demo-data button. First analyze if the user prompt describes specific look & feel requirements. For instructional text and demo data: include them only when the app is clearly a plain CRUD app without specified look & feel. Read the JSON payload and return JSON with properties: "selected" (array of catalog "name" strings), "instructionalText" (boolean), and "demoData" (boolean). Only choose modules from the catalog. Include any libraries already used in history. Respond with JSON only.',
     },
     { role: 'user', content: JSON.stringify(payload) },
   ];