How datadog-labs/agent-skills works
This repository is a collection of Datadog skills that guide AI agents through product workflows; it is not primarily a runtime service. The usual path is an engineer asks an agent for a task, the relevant skill supplies procedural guidance, and the agent uses Datadog tools or local project context to produce a result. README documents the main skill capabilities and an explicit Agent Observability pipeline. The graph uses those documented paths and relationships; source excerpts are limited, so implementation-level links between other skills are omitted.
Main entry points: agent-observability-eval-pipeline/SKILL.md Eval Pipeline · agent-observability-session-classify/SKILL.md Session Classification · dd-pup/SKILL.md Pup CLI · dd-monitors/SKILL.md Monitor Management
Components 27
Core Datadog
- Pup CLI
dd-pup/SKILL.md - Monitor Management
dd-monitors/SKILL.md - Log Search
dd-logs/SKILL.md - APM Guidance
dd-apm/SKILL.md - Documentation Search
dd-docs/SKILL.md
Agent Observability
- Session Classification
agent-observability/agent-observability-session-classify/SKILL.md - Trace RCA
agent-observability/agent-observability-trace-rca/SKILL.md - Evaluator Bootstrap
agent-observability/agent-observability-eval-bootstrap/SKILL.md - Eval Pipeline
agent-observability/agent-observability-eval-pipeline/SKILL.md - Experiment Analysis
agent-observability/agent-observability-experiment-analyzer/SKILL.md - Experiment Bootstrap
agent-observability/agent-observability-experiment-bootstrap/SKILL.md - Annotation Evaluator
agent-observability/agent-observability-build-eval-from-annotations/SKILL.md - Prompt Optimization
agent-observability/agent-observability-auto-experiment/SKILL.md - Trace Replay
agent-observability/agent-observability-replay-trace/SKILL.md
Application Workflows
- RUM Instrumentation
dd-instrument-rum/SKILL.md - Browser SDK
dd-browser-sdk/SKILL.md - Datadog Apps
dd-apps/datadog-app/SKILL.md - Product Recommender
dd-product-recommender/SKILL.md
Security and Delivery
- Audit Investigations
dd-audit/SKILL.md - Security Ownership
dd-security/csm/ownership-agent/SKILL.md - Delivery Workflows
dd-software-delivery/unblock-pr/SKILL.md - Cloud Integrations
dd-aws-integration/SKILL.md
Other components
- Engineer
- AI Agent
- Datadog Platform
- RUM Signals
- Local Application
Connections 43
- Engineer → AI Agent: requests task
- AI Agent → Pup CLI: uses guidance
- AI Agent → Monitor Management: uses guidance
- AI Agent → Log Search: uses guidance
- AI Agent → APM Guidance: uses guidance
- AI Agent → Documentation Search: uses guidance
- Pup CLI → Datadog Platform: authenticates and operates
- Monitor Management → Datadog Platform: manages monitors
- Log Search → Datadog Platform: searches logs
- APM Guidance → Datadog Platform: investigates traces
- Documentation Search → Datadog Platform: searches documentation
- AI Agent → Session Classification: uses guidance
- AI Agent → Trace RCA: uses guidance
- AI Agent → Eval Pipeline: uses guidance
- AI Agent → Experiment Analysis: uses guidance
- AI Agent → Experiment Bootstrap: uses guidance
- AI Agent → Annotation Evaluator: uses guidance
- AI Agent → Prompt Optimization: uses guidance
- AI Agent → Trace Replay: uses guidance
- Eval Pipeline → Session Classification: classifies sessions
- Eval Pipeline → Trace RCA: diagnoses failures
- Eval Pipeline → Evaluator Bootstrap: builds evaluators
- Eval Pipeline → Datadog Platform: publishes and runs evaluations
- Session Classification → RUM Signals: combines behavioral signals
- AI Agent → RUM Instrumentation: uses guidance
- AI Agent → Browser SDK: uses guidance
- AI Agent → Datadog Apps: uses guidance
- AI Agent → Product Recommender: uses guidance
- RUM Instrumentation → Datadog Platform: sends RUM data
- Browser SDK → Datadog Platform: guides SDK telemetry
- Datadog Apps → Datadog Platform: queries platform data
- AI Agent → Audit Investigations: uses guidance
- AI Agent → Security Ownership: uses guidance
- AI Agent → Delivery Workflows: uses guidance
- AI Agent → Cloud Integrations: uses guidance
- Audit Investigations → Datadog Platform: investigates audit data
- Security Ownership → Datadog Platform: uses security context
- Delivery Workflows → Datadog Platform: uses MCP and Pup
- Trace Replay → Local Application: replays trace against
- Prompt Optimization → Local Application: scores code changes
- Experiment Analysis → Datadog Platform: analyzes experiments
- Experiment Bootstrap → Datadog Platform: uses LLM observability SDKs
- Annotation Evaluator → Datadog Platform: publishes evaluator draft