getsentry/sentry-for-claude
Overview
This skill analyzes and resolves Sentry review comments left on GitHub pull requests. It locates Sentry bot comments, extracts the bug description, analysis, suggested fix, and AI-agent prompt, verifies the issue in the code, and applies precise fixes when appropriate. Use it to turn Sentry feedback into concrete, safe code changes and a clear resolution report.
How this skill works
Given a PR number or URL, the skill fetches line-specific review comments and filters only comments authored by Sentry (usernames that start with "sentry"). It parses the comment body to extract the Bug header, severity/confidence, detailed analysis, suggested fix, and AI prompt. For each comment it reads the target file and line, verifies whether the issue still exists, implements a targeted fix if safe, and produces a summary report of resolved and outstanding items.
When to use it
- You receive Sentry line-specific review comments on a GitHub pull request and want automated analysis and fixes.
- You want a consistent process to verify Sentry-reported issues before making code changes.
- You need a concise report summarizing which Sentry issues were fixed and which need manual review.
- You want to batch-process recent PRs to find and address Sentry feedback automatically.
Best practices
- Only process comments whose author starts with "sentry"; ignore other bots and human comments.
- Always read the file and the exact line referenced before making edits—verify the problem first.
- Prefer minimal, precise changes that address the root cause and consider edge cases.
- When unsure about side effects or safety, pause and ask the user for guidance before editing.
- Run the project's test suite and linting after applying fixes to catch regressions.
Example use cases
- A PR receives multiple Sentry comments about missing null checks; use the skill to add guards and report fixes.
- Sentry flags a permissive validation path; parse the suggestion, tighten validation, and provide a patch summary.
- Sentry reports an error-handling gap where exceptions are swallowed; implement re-throwing or proper logging and mark resolved.
- Scan recent PRs in a release branch to discover any Sentry feedback tagged as production-impacting and triage them.
FAQ
It only processes review comments where the author username starts with "sentry"; all other bots and human comments are ignored.
Will the skill change code without confirmation?
The skill verifies issues before editing and will apply precise fixes when confident; if uncertain it requests user guidance.