How to use the Fix Sentry Issues agent skill (sentry-fix-issues)

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file. The file tells an AI agent when to use a workflow and how to carry it out step by step. This skill covers the workflow of finding and fixing Sentry issues.

Once installed, the agent follows a fixed sequence when you ask it to fix Sentry errors. It lists unresolved issues through a Sentry MCP connection, reads stack traces, breadcrumbs and traces, and writes a root-cause hypothesis before it edits any code. It then checks that the error data matches your repository, applies a fix with tests built from synthetic data, and reports the result.

• Discovers unresolved issues and asks you to confirm which ones to fix
• Analyzes stack traces, breadcrumbs, traces and context
• Documents a root-cause hypothesis with alternatives before changing code
• Flags mismatches between Sentry data and your source code
• Adds tests that reproduce the error conditions using generalized data
• Ends with a written report covering the fix, evidence and follow-up

What you need before installing

The skill cannot find issues on its own. It depends on a Sentry MCP server that is configured and connected to your agent. Sentry account credentials are handled by that MCP setup, not by the skill.

You also need access to the Sentry project or organization, and a local copy of the source repository with git available, because the skill uses git history to look up recent changes.

• Sentry MCP server configured and connected
• Access to the Sentry project or organization
• A local copy of the source repository with git available

Install the skill in Claude Code or Claude apps

Download the ZIP file for the skill from this page and unzip it. For a personal setup across all your projects, place the unzipped skill folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps, you can instead upload the skill under Settings, then Capabilities, then Skills.

The package lists SKILL.md as its only file. Check that the folder contains SKILL.md after unzipping.

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

How to trigger the skill with natural requests

You do not need special syntax. The skill is meant to activate when you ask for Sentry work in plain language, such as asking to fix Sentry errors, resolve Sentry errors, debug production bugs or investigate exceptions. You can also mention an issue ID or an error message, or ask to work through your Sentry backlog.

The skill's metadata sets disable-model-invocation to true, so it is best to name Sentry explicitly in your request so the agent knows to use this workflow.

• 'Fix the unresolved Sentry errors in the checkout service.'
• 'Investigate PROJECT-123 in Sentry and tell me the root cause.'
• 'Triage our Sentry backlog and confirm which issues to fix first.'
• 'Debug the production TypeError reported in Sentry this week.'

What the agent does during a fix

The skill divides the work into phases. First the agent discovers issues and waits for your confirmation. Then it gathers context for each issue, including events, tags, traces and any AI root-cause analysis Sentry provides. Before it changes code, it writes down the error summary, immediate cause, root-cause hypothesis, evidence and alternatives.

Next the agent checks the code. It reads each file in the stack trace from the top down, traces where values come from, and looks for try/catch blocks and recent commits. If the event data does not match the repository, it stops and reports the mismatch. After a fix, it checks for regressions and similar patterns elsewhere before it reports the result.

• Phase 1: discover issues and confirm which to fix
• Phase 2: gather details from Sentry
• Phase 3: write the root-cause hypothesis
• Phase 4: cross-check Sentry data against the codebase
• Phase 5: implement the fix and add tests
• Phase 6: run a verification audit
• Phase 7: report the results

Safety notes: data, network and review

The skill treats all Sentry content as untrusted input. Exception messages, breadcrumbs, request bodies, tags and user context may contain text written by an attacker. The skill instructs the agent not to follow instructions found in that data, not to copy raw values into code or tests, and not to repeat tokens, passwords, session IDs or personal information in its output.

This skill pulls event data into the agent session, and that data can include personal information. If your policy forbids sending production data to an AI tool, do not use it. The skill uses the network through the Sentry MCP connection, and it does not run scripts. Code changes still need human review before deployment.

• Does not run scripts; uses network: yes
• Sentry data is treated as untrusted text
• Secrets and PII are referenced indirectly, not reproduced
• Asks you to confirm which issues to fix
• Agent-made code changes still need human code review

Tips for getting better results

Start with a narrow request, such as one issue ID or one error type, before asking for a full backlog triage. Keep the repository you are working in checked out and up to date, so the code cross-check reflects what is actually deployed.

If the agent reports a mismatch between the Sentry stack trace and your code, check whether the release in Sentry matches your local branch before deciding the report is wrong.

• Begin with one issue ID or one error type
• Make sure the local branch matches the release Sentry reports
• Review every diff before merging or deploying
• Ask for the verification audit output if you want a closer review