How to use the Sentry Error Debugging agent skill

What the Sentry skill makes the agent do

An agent skill is a set of instructions that an AI agent reads to change how it approaches a task. This one covers Sentry error tracking. Once installed, the agent follows the practices in the skill when you ask about Sentry: setting up the SDK, sorting new errors by real user impact, reading stack traces and breadcrumbs, and configuring alerts that avoid noise.

The skill only supplies guidance. It does not connect to your Sentry account, and it cannot fetch issues or read live data.

• SDK setup: initialize early, set environment and release on every event, and scrub personal data with beforeSend or before_send.
• Triage: review new unresolved issues, check frequency and user impact, read the stack trace, then check breadcrumbs, tags and context.
• Alerts: alert on new issue types and error spikes, use issue.priority and event.frequency conditions, and route alerts to the right team.
• Performance: use distributed tracing, set thresholds per transaction type, and look for N+1 queries and slow database spans.

Install the Sentry skill with a ZIP download

Download the ZIP file for the skill from this page and unzip it. The skill is a folder that contains a single file, SKILL.md.

For a personal install, unzip the folder into ~/.claude/skills/. To limit the skill to one project, unzip it into that project's .claude/skills/ folder instead.

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: the .claude/skills/ folder inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

Trigger the skill with natural requests

You do not need a special command. Describe what you need in plain language that mentions Sentry, errors, alerts or performance, and the agent can use the skill. Give it the context it needs, such as the language, framework and what you have already tried.

The skill's guidance is most useful when you describe a concrete situation. Pasting a stack trace you have already scrubbed of personal data works well.

• Set up Sentry in a Node.js API and tell me where to initialize it.
• Help me triage the unresolved issues from the last 24 hours and rank them by user impact.
• Our alerts fire all night. Suggest conditions to cut the noise.
• Read this stack trace and tell me which function failed and what input likely caused it. [scrubbed trace pasted]

Examples of what the agent will recommend

Here is the kind of guidance the skill contains. The agent applies these practices to your specific situation.

For triage, the skill suggests reviewing new issues with the Issues page filtered by is:unresolved firstSeen:-24h. It also says a rare error in a critical path can be worse than a frequent one in a niche feature.

For SDK setup, it recommends setting traces_sample_rate by traffic volume: 1.0 for low-traffic services and 0.1 to 0.01 for high-traffic ones.

What the skill needs and what it does not do

The skill lists no requirements. It runs no scripts and uses no network access, according to its own listing. It is a single SKILL.md file of instructions.

It does not integrate with the Sentry API. Anything the agent says about your live issues comes from what you paste into the chat, not from your account. Your own Sentry project, SDK and any ticketing tools still need to be set up by you.

Safety notes before you share data

The skill itself warns that personal data should never reach Sentry. Before you paste error logs or user data into the agent, remove names, emails, IP addresses, passwords and auth tokens. The agent may echo sensitive content back, and anything you paste is shared with the agent in that conversation.

Use the scrubbing hooks described in the skill, beforeSend or before_send, so personal data is removed from events before they are sent to Sentry.

• Scrub logs and traces before pasting them into the chat.
• Configure beforeSend or before_send to remove PII at the source.
• Do not auto-resolve issues just to clear a queue; they will return.

Tips for getting better results

Tell the agent which Sentry features you use, such as performance monitoring or source maps, so it focuses on the right guidance. Mention your traffic level so it can suggest a sensible sample rate.

Ask for one task at a time. Triage, alert tuning and SDK setup each have their own checklist in the skill, and focused requests give clearer answers.

Keep the skill's pitfalls in mind. The skill warns against ignoring rate limits, since critical errors may be dropped once you exceed your quota, and against running a 100% trace sample rate on high-traffic services.

• Name your language or framework and environment (production, staging, development).
• Paste short, scrubbed excerpts rather than full logs.
• Ask the agent to explain its reasoning so you can check it against your own data.