agentleFS
datadog-labsdatadog-labs/agent-skills/dd-instrument-error-tracking/SKILL.md

dd-instrument-error-tracking

Skill185 starsChanged 38 days agoMITView on GitHub

Sets up Datadog Browser Error Tracking for a web app that runs the Datadog RUM Browser SDK. Turns error processing on for the app's RUM application, forwards framework-caught errors (React, Next.js, Vue, Angular, SvelteKit) to Datadog, and checks that events arrive. Use when asked to add, set up, enable or verify Error Tracking or frontend exception tracking in a browser app. Not for adding RUM itself, which dd-instrument-rum does first, or for uploading source maps, which dd-browser-sdk does.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Two or more inits, or npm and CDN setups mixed: stop and report each location, because a second init splits sessions and errors between …

    line 33
  2. Distinctive:

    No application ID found: ask the user. It is listed under Real User Monitoring > Manage Applications in Datadog.

    line 34
  3. Very specific:

    If the existing datadogRum.init sets sessionSampleRate: 0, no sessions are sampled and no errors are sent.

    line 39
  4. Very specific:

    If @datadog/browser-logs is already installed and datadogLogs.init is present without forwardErrorsToLogs, add forwardErrorsToLogs: true.

    line 40
  5. Very specific:

    Leave the credentials, site, service, and unrelated options in datadogRum.init unchanged.

    line 41

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Datadog Browser Error Tracking

Browser Error Tracking groups and triages frontend exceptions collected by the Datadog RUM Browser SDK (@datadog/browser-rum). It shares the same SDK and the same RUM application as Browser RUM and Product Analytics, so there is no separate Error Tracking SDK to install.

How it works

  • The switch is on the RUM application, not in the code. A RUM application processes errors when rum_event_processing_state is ERROR_FOCUSED_MODE or ALL. An application created for Product Analytics only starts at NONE, which drops all RUM and Error Tracking events even when the browser code is wired.
  • RUM comes first. Browser Error Tracking reads what the RUM Browser SDK sends, so an app with no RUM init needs RUM set up first.
  • Sampling uses sessionSampleRate. The RUM Browser SDK captures unhandled errors (window.onerror, unhandledrejection, console.error) automatically for sampled sessions. There is no errorSampleRate option in RumInitConfiguration; adding it breaks TypeScript builds.
  • Frameworks swallow component errors. React error boundaries, Next.js error.tsx, Vue app.config.errorHandler, Angular ErrorHandler, and SvelteKit handleError intercept errors before they reach window.onerror, so those hooks must forward caught errors to Datadog.
  • Credentials. The application ID and client token are public and already in the RUM setup. Switching event processing on through the API needs an API key and an application key, and both are secret.

Set up

Copy these steps into your reply and tick them off as you go.

  1. Inspect. Find the browser app; in a monorepo with several, ask which one. Search source and HTML files, skipping node_modules and build output, for the RUM setup: an init call on the datadogRum import from @datadog/browser-rum, including renamed imports and wrapper modules, or a CDN DD_RUM.init. Record the number of inits, the framework, sessionSampleRate, the site (absent means datadoghq.com), the application ID, any @datadog/browser-rum-* or @datadog/browser-logs packages, and any existing error boundary or framework error handler. When the init reads the application ID or the site from environment variables, read only those lines of the env file, for example grep -E '^VITE_DD_(RUM_APPLICATION_ID|SITE)=' .env.local, because env files often hold secrets.
  2. Check what Error Tracking needs.
    • No browser app, for example a backend service or a mobile app: stop before any edit and report that this skill covers browser apps only. Do not offer Browser RUM, because it cannot collect errors from that project.
    • No RUM init: stop before any edit. Tell the user that Browser Error Tracking collects its data through the RUM Browser SDK, that this app has none, and that RUM must be set up first with the dd-instrument-rum skill. Ask whether to set up RUM now. On yes, run that skill; when it is not installed, get it from that link or follow the Browser Monitoring setup docs. Then restart at step 1, and if there is still no init with an application ID, stop and report it. On no, or when you cannot ask, end without changes and report that RUM is missing.
    • Two or more inits, or npm and CDN setups mixed: stop and report each location, because a second init splits sessions and errors between configurations. Repairing that belongs to RUM setup.
    • No application ID found: ask the user. It is listed under Real User Monitoring > Manage Applications in Datadog.
  3. Turn Error Tracking processing on. Check for both keys without printing them: [ -n "$DD_API_KEY" ] && [ -n "$DD_APP_KEY" ] && echo set. With keys, follow references/api-switch.md: it reads the current state, leaves ALL or ERROR_FOCUSED_MODE untouched, and sends rum_event_processing_state: ERROR_FOCUSED_MODE only when the current state is NONE. Without keys, or after a failure, give the user the settings step: in Datadog, open Digital Experience > Real User Monitoring > Manage Applications, select the application, and under Application Settings enable Error Tracking event processing. Build APP_HOST as described in Verify and view. When the Datadog MCP server has the RUM toolset, Datadog:search_rum_applications reports rum_event_processing_state, which confirms either path.
  4. Check session sampling and Logs forwarding.
    • If the existing datadogRum.init sets sessionSampleRate: 0, no sessions are sampled and no errors are sent; ask before changing it to 100. When sessionSampleRate is absent, leave it alone (it defaults to 100). Never add errorSampleRate.
    • If @datadog/browser-logs is already installed and datadogLogs.init is present without forwardErrorsToLogs, add forwardErrorsToLogs: true. Do not install @datadog/browser-logs from scratch, because Browser Logs is a separate product with its own setup.
    • Leave the credentials, site, service, and unrelated options in datadogRum.init unchanged.
  5. Forward framework-caught errors. Open references/framework-errors.md and wire the error hook for the detected framework (React, Next.js, Vue, Nuxt, Angular, or SvelteKit) so errors caught by the framework are reported to Datadog without replacing existing UI or reporting the same error twice. Plain JavaScript, plain Svelte, and CDN apps need no framework hook.
  6. Build and run. Run the project's build, fix errors your edits caused, and run it again until it passes. Then start the app with its dev command, load a page, keep the page open for about 30 seconds, and stop the server. The SDK sends its batch every 30 seconds or when the page is hidden.

Verify and view

Record the highest level you actually reached:

  1. Configuration changed: the switch result from step 3 and the code edits are in place.
  2. The app ran: the project built cleanly and a page loaded in a browser.
  3. Export accepted: a request to a browser-intake-… host, or to the init's proxy URL, with path /api/v2/rum, returned 202. Check in a headless browser when one is available, or ask the user to look in DevTools > Network.
  4. Datadog shows events: Datadog:search_datadog_rum_events, or the Search RUM events endpoint with the same keys, finds @application.id:<APP_ID> @session.initial_view.url_host:localhost in the last 15 minutes. A new application can take a few minutes to show data, so query once a minute for up to 10 minutes; if nothing arrives, stop at level 3 and give the link.

A passing build shows only that the project compiles. Name each level you could not check and the step left for the user.

The app host is app.datadoghq.com for datadoghq.com, app.datadoghq.eu for datadoghq.eu, app.ddog-gov.com for ddog-gov.com, and the site itself for the others, for example us5.datadoghq.com. End with this summary:

Error Tracking: turned on | already on | waiting for your settings step (<the step>)
Changed: <files, or none>
Evidence: <highest level reached>; not checked: <levels and the step left>
View: https://APP_HOST/error-tracking?query=@application.id:<APP_ID>

Undo

Revert this run's code edits (framework error hook or boundary). If this run changed rum_event_processing_state from NONE to ERROR_FOCUSED_MODE, follow the undo note in references/api-switch.md. Leave the RUM setup as it is.

Gotchas

MistakeSymptomFix
Sending rum_event_processing_state: ERROR_FOCUSED_MODE when the app is at ALLDowngrades full RUM and drops non-error sessionsRead the state first; update only when it is NONE
Sending product_analytics_retention_state: NONE in the updateTurns Product Analytics off for the appOmit product_analytics_retention_state from the PATCH body
Adding errorSampleRate to datadogRum.initTypeScript build fails (errorSampleRate is not a valid option)Use sessionSampleRate (defaults to 100)
Calling datadogRum.addError in server-only Next.js or SvelteKit codeServer error is never sent by the browser SDKCall it in a client error boundary ('use client') or hooks.client

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.