Establish agent skills as source of truth, simplify example content

[pkosiec]

Summary

Companion to databricks-agent-skills PR #76.
⚠️ databricks/databricks-agent-skills#76 should be merged first!

Establishes Databricks agent skills as the single source of truth for implementation patterns, while keeping DevHub templates as the discovery and context layer.

What changes for agents

• Preamble: New "Implementation guidance" section in dev-guidelines.md tells agents the template describes what to build, and installed skills provide the how.
• Intent blocks: Each intent file (recipe, cookbook, example) now instructs agents to treat skills as the source of truth when template content and skills overlap.
• Example prompts: Simplified buildFullPrompt() — removed duplicate CLI auth verification steps that the local-bootstrap block already covers.

What changes for users

• Example pages: inventory-intelligence content simplified to outcome-only (Data Flow + What to Adapt). The "Setup — interview the user" section is removed (see below).
• Cookbook intro: ai-chat-app intro now mentions that installed skills provide implementation patterns.
• Recipe pages: No content changes. Recipes still serve humans browsing the site with full implementation details.

Why we removed the "interview the user" pattern

The inventory-intelligence example had a structured 8-question interview (workspace URL, CLI profile, catalog, warehouse ID, Lakebase project, data mode, Genie tab, forecast model) baked into content.md. This is no longer needed because:

1. The intent blocks (intent-example.md) already define the conversational flow — clarify intent, pin down decisions, ask one question at a time
2. The skills individually prompt for the pieces they need (profile name, Lakebase project, etc.) as part of their own workflows
3. The example's template/README.md has the actual setup instructions

With skills as the source of truth, agents discover what to ask as they go — they don't need a pre-baked questionnaire.

Test plan

• npm run build passes
• npm run test passes (255/255)
• npm run typecheck passes
• Example pages render correctly (spot-check /templates/inventory-intelligence)
• Agent prompt includes "Implementation guidance" preamble (spot-check /api/markdown)
• Cookbook intro reads naturally (/templates/ai-chat-app)