A Microsoft-native adaptation of context engineering skills for production agent systems. Use when building, optimizing, evaluating, governing, publishing, or debugging agent systems with core context-engineering mechanisms plus Azure AI and Microsoft Foundry…
This skill should be used when publishing Microsoft-native agents: Foundry agent versions, stable endpoints, Teams and Microsoft 365 Copilot distribution, Entra Agent Registry, environment promotion, rollout, and rollback.
This skill should be used when implementing private grounded retrieval on Azure: Azure AI Search, agentic retrieval, knowledge sources, File Search, SharePoint grounding, citations, access-controlled indexes, and retrieval traces.
This skill should be used when binding agent access to Microsoft Entra identity: managed identities, RBAC scopes, on-behalf-of access, tenant boundaries, credential-free SDK use, and per-user authorization for tools and grounding sources.
This skill should be used when implementing Microsoft-native agent memory and state: Cosmos DB, Blob or ADLS scratchpads, Azure Cache for Redis, session stores, episodic memory, entity state, vector memory, and append-only logs.
This skill should be used when grounding agents in Microsoft Fabric: Fabric Data Agent, semantic models, lakehouses, warehouses, Power BI aligned metrics, governed analytical data, and natural-language data access.
This skill should be used when designing Microsoft Foundry Hosted Agents and managed agent runtimes: production endpoints, runtime boundaries, sandboxing, scaling, tool execution, runtime state, deployment environments, and operational controls.
This skill should be used when designing Microsoft Foundry IQ knowledge bases, reusable enterprise grounding, federated knowledge APIs, permission-aware retrieval, knowledge source governance, and shared context layers for agents.