What problem does it solve?
Crash reports arrive as scattered notes, Gmail exports, or Sentry links, and turning them into verified root-cause fixes requires manual triage, repository searching, and validation. This Skill walks an AI agent through the full loop: reading the note, fetching live Sentry context, diagnosing the crash, deciding whether it is worth fixing, implementing the fix, validating it, and optionally committing and resolving the Sentry issue.
Core Features & Use Cases
- Sentry context retrieval: Extracts issue IDs, event IDs, releases, and stack frames from notes or URLs, then fetches live issue and event data through Sentry MCP tools or CLI helpers.
- Root-cause diagnosis and impact triage: Identifies the first-party frame versus the throwing library frame, weighs occurrence counts and affected users, and explicitly supports skipping low-impact or noise reports without speculative fixes.
- Guided fix and validation: Searches the repo with rg, checks dependency versions in gradle/libs.versions.toml, applies narrow patches, and runs targeted Gradle builds, detekt, and tests before handoff.
- Use Case: A user pastes a Sentry issue URL for a crash in the FeedFlow desktop app and asks for a fix; the Skill fetches the event, traces the crash to a dependency API change, patches the call site, runs :desktopApp:compileKotlinJvm, and resolves the Sentry issue with the commit reference.
Quick Start
Use $sentry-note-crash-fix to investigate and fix the crash described in this Sentry issue URL.