We need to stop treating AI as a tool. It's time to treat it as a peer.
I've been building a library of reusable skills for Claude — structured instructions that let AI agents handle complex, repetitive development workflows on Databricks and Azure AI Foundry.
The goal is simple: stop re-explaining the same context every time, and stop reinventing the wheel.
Instead of building everything from scratch, each skill is designed to compose — pulling in existing work from AI Dev Kit, LangGraph and LangChain skill ecosystems, then extending them with Databricks and Azure-specific context. Load a skill and Claude operates with the same domain knowledge a senior engineer would bring to the table: the right patterns, the right APIs, the right tradeoffs — already loaded.
So far the library covers:
- Databricks Mosaic AI Agents — Model Serving, LangGraph, LangChain, OpenAI Agent SDK, Deep Agents, Human-in-the-Loop, Supervisor Agents
- Azure AI Foundry Agents — SDK v2, Memory, Durable Orchestration, AI Gateway, observability with App Insights.
Each skill is versioned, documented, and installable. The repo also includes an MCP server layer, for ai-dev-kit, so these capabilities can be reused across multiple workflows.
But beyond the technical side, this project is a statement.
We spend a lot of energy debating whether AI will replace developers. That's the wrong question. The right question is: are you using AI to multiply what you can build?
When I treat Claude as a peer — giving it context, structure, and clear objectives — I move faster, think more clearly, and ship better work. Not because I'm doing less, but because I'm doing more of
what actually matters.
The shift isn't about automation. It's about raising the floor of what one engineer can do.
The repo is open. If you're building on Databricks or Azure AI and want to skip the boilerplate, it might be worth a look.
https://github.com/alessandro9110/databricks-foundry-agent-skills
