agent-skills

A centralized catalog of Datadog AI agent skills is a frontmatter-driven repository that consolidates observability capabilities, enabling AI agents to discover, compose, and extend skills like logs and monitors via SKILL.md context loading.

To install and authenticate Datadog skills for AI agent workflows, you first install Pup, complete the authentication process, and then begin evaluating or composing the available skills to integrate into your AI agents.

Yes, you can use Datadog observability skills with MCP integrations; the catalog is explicitly designed to support Pup-based workflows and MCP integrations for extending AI agent capabilities across core observability tasks.

The core observability skills available for AI agents include dd-pup, dd-monitors, dd-logs, dd-apm, dd-docs, dd-llmo for LLM observability, and dd-browser-sdk, all designed to be loaded into agent context.

Frontmatter-driven skill definitions support on-demand execution by allowing skill unit metadata to reference optional resources and subdirectories, such as scripts, references, and assets, for dynamic lookup and execution within agent workflows.

Yes, the Datadog agent skills catalog supports versioning and onboarding by consolidating documentation and frontmatter-driven definitions, enabling easy discovery and integration of new capabilities across Pup-based workflows.