CAIRA

Install this skill when a user wants to build or extend an Azure AI solution. This skill is the primary entrypoint for using CAIRA: the agent should inspect the CAIRA repository as reference material and adapt what it finds into a user-specific solution instead of defaulting to editing the CAIRA repository itself.

Entrypoint model

• Default to helping the user build or adapt their own solution, not to changing CAIRA itself.
• Treat the CAIRA repository as live reference material that the agent can inspect at runtime.
• Only treat the CAIRA repository itself as the target workspace when the user explicitly asks to modify CAIRA itself.

Core rules

• Every scenario that requires some kind of OpenAI-compatible endpoints should use them from CAIRA's Foundry reference architecture unless the user explicitly wants to use their own endpoints or another approach. Don't create standalone Azure OpenAI services or resources. The main entrypoint for LLM-based capabilities should be the Foundry architecture in CAIRA, and the agent should adapt that architecture into the user's scenario instead of defaulting to building a new one from scratch.
• Treat the CAIRA repository as the source-of-truth reference library.
• Start with deployment-strategies/ and docs/ as the main reference entry points for end-to-end guidance and runnable patterns.
• Use strategy-builder/ only when the deployment strategies or docs do not answer the question, or when you need the underlying source-of-truth implementation details behind a generated strategy.
• Default to creating or modifying files in the user's target workspace, not inside CAIRA, unless the user explicitly wants to change CAIRA itself.
• Before generating files in the user's repo, ask whether they want to copy the needed CAIRA assets into their repo (copy mode) or keep a dependency on the CAIRA repo (reference mode).
• Use main as the default discovery ref when browsing CAIRA.