datadog - SKILL.md Agent Skill

name: datadog description: Use Datadog MCP tools to investigate logs, metrics, traces, and incidents for the Gram project. Activate when the user asks about errors, performance issues, incidents, latency, or wants to search telemetry data.

Always filter by the relevant service(s) when querying Datadog:

Service	Description
`gram`	Dashboard frontend (RUM)
`gram-dashboard`	Dashboard backend
`gram-server`	Main backend API server
`gram-worker`	Temporal worker
`fly`	Fly.io — where Gram Functions logs are collected

Use only the following Datadog MCP tools unless the user explicitly asks for others:

mcp__datadog-mcp__search_datadog_logs — Search and filter log events
mcp__datadog-mcp__analyze_datadog_logs — Analyze log patterns and aggregate stats

mcp__datadog-mcp__get_datadog_metric — Get a specific metric's time-series data
mcp__datadog-mcp__get_datadog_metric_context — Get context and metadata for a metric
mcp__datadog-mcp__search_datadog_metrics — Search available metrics by name

mcp__datadog-mcp__get_datadog_trace — Get a specific trace by ID
mcp__datadog-mcp__search_datadog_spans — Search spans (useful for latency investigation)

mcp__datadog-mcp__search_datadog_incidents — Search active/recent incidents
mcp__datadog-mcp__get_datadog_incident — Get details for a specific incident
mcp__datadog-mcp__search_datadog_monitors — Find monitors and their current state

mcp__datadog-mcp__search_datadog_rum_events — Search Real User Monitoring events (frontend errors, sessions)
mcp__datadog-mcp__search_datadog_events — Search Datadog events stream

mcp__datadog-mcp__search_datadog_services — Discover services in APM
mcp__datadog-mcp__search_datadog_service_dependencies — View service dependency map

Always scope queries to one or more Gram services using the service filter when available.
Start narrow, expand if needed: Query a 15–30 minute window first, then widen.
For error investigations: start with search_datadog_logs, filter by status:error, then follow trace IDs with get_datadog_trace.
For latency issues: use search_datadog_spans with service:<name> and sort by duration.
For frontend issues: prefer search_datadog_rum_events for gram.
For incidents: check search_datadog_incidents first before deep-diving into logs.