datadog-docs

Skill: datadog-docs

Execute Datadog API requests to fetch live objects, metrics, or monitor statuses, and retrieve Datadog documentation via LLM-optimized endpoints.

When to Use

• Documentation Search: Locating specific Datadog documentation articles or reference materials.
• Article Retrieval: Fetching Datadog documentation using the provided LLM-optimized endpoints.
• When querying Datadog metrics, logs, or traces directly from the command line in an automated script.
• To retrieve and analyze the status of Datadog monitors.
• When you need to fetch a dashboard configuration JSON to inspect or backup its definition.

When Not to Use

• When full MCP observability tools (datadog-mcp) are available, as they natively format data for LLM context windows.
• If the project uses Pulumi or Terraform for Datadog resource creation (use the respective Infrastructure-as-Code skill).
• When a simple web UI link to the Datadog dashboard would suffice for a human user.
• When you need to configure or install the Datadog Agent. Use datadog-agent instead.

Common Pitfalls

• Paginating Endlessly: Running an API query that returns millions of log lines without using jq to truncate or filter the output, causing the agent to run out of memory.
• Leaking API Keys: Accidentally echoing the DD-API-KEY or DD-APPLICATION-KEY in the terminal during debugging, exposing it in the workflow logs.
• Incorrect Endpoint Region: Sending requests to datadoghq.com when the organization's Datadog instance is actually hosted on datadoghq.eu or us3.datadoghq.com, resulting in HTTP 403 or 404 errors.

Core Process

1. Identify the Need: Determine what Datadog documentation or API resource is needed.
2. Access API Reference: Check references/api_reference.md for the complete catalog of API endpoint documentation links.
3. Verify Credentials (API calls): Ensure DD-API-KEY and DD-APPLICATION-KEY are available in the environment.
4. Determine Datadog Site: Default is datadoghq.com, but it could be datadoghq.eu, us3.datadoghq.com, etc. based on the target environment.
5. Access Documentation (docs lookup): Use the provided llms.txt to access broader documentation links.
6. Execute API Call (API interaction): Use curl to construct the API request, passing necessary headers and query parameters.
7. Process Response: Use tools like jq to parse and extract relevant data from the JSON response.

Core Principles

• Secure Headers: Always pass keys via headers, never in the URL path.
• Data Pruning: Use jq or python to prune large JSON responses, especially when fetching dashboards or monitors.
• Pagination: Handle pagination when querying lists of items using start or page parameters.

References

• API Reference MUST be fetched when interacting with the Datadog API.
• Datadog API cURL Examples MUST be fetched when constructing cURL requests for the Datadog API.
• Datadog MCP (Bits AI) MUST be fetched when querying or analyzing Datadog telemetry via MCP.
• Datadog Monitors MUST be fetched when designing, debugging, or troubleshooting Datadog monitor queries.
• Datadog Agent MUST be fetched when configuring, installing, or troubleshooting the Datadog Agent.
• https://docs.datadoghq.com/llms.txt - The official Datadog documentation curated for LLMs. MUST be fetched for the full context.
• https://docs.datadoghq.com/monitors/llms.txt - The official Datadog Monitors documentation curated for LLMs. MUST be fetched when designing, debugging, or interacting with monitors.
• https://docs.datadoghq.com/agent/llms.txt - Install and configure the Agent to collect data. MUST be fetched when configuring, installing, or troubleshooting the Datadog Agent.
• Datadog API Documentation: https://docs.datadoghq.com/api/latest/ MUST be fetched when looking up specific API endpoints missing from local references.

Diagnostics and Troubleshooting

• 403 Forbidden or Failed permission authorization checks: Check DD-API-KEY and DD-APPLICATION-KEY validity and ensure they match the correct Datadog site (datadoghq.com vs datadoghq.eu). Also verify that the Application Key has the required scopes for the targeted API endpoint.
• 429 Too Many Requests: Implement exponential backoff if hitting API rate limits.
• 404 Not Found: Verify the ID, endpoint path, and ensure you are hitting the correct region endpoint.

What to Avoid

• Avoid logging or exposing DD-API-KEY or DD-APPLICATION-KEY in clear text outputs.
• Do not fetch massive metric datasets without strict from and to bounds to avoid overloading the context window.

Limitations

• The agent context window can be easily overwhelmed by full dashboard JSONs. Always filter the response with jq when possible.
• For retrieving telemetry (logs, metrics, traces, monitors), always prefer optimized MCP-based tools over raw API calls if available, as they are specifically designed to optimize data size for the agent's context window.