crash-fixer
by ryce
DevOps & CloudBrief
Autonomous crash analysis and bug fixing.
Description
What This Skill Does
Fetches crash reports, deduplicates by fingerprint, analyzes each crash with AI, generates a code fix, and opens a pull request automatically. Supports dry-run mode to preview analysis without creating PRs. Configurable by time window and per-run crash limit.
Closes the full loop from crash detection to merged fix without manual triage, deduplication, or PR creation.
When to Use It
- Automatically patching overnight production crashes before the team starts work
- Running dry-run analysis during an incident to understand root causes without committing
- Capping PR noise by processing only the 3 most recent crashes per scheduled run
- Checking the last 2 hours of crashes after a deploy to catch regressions quickly
- Letting a CI job handle crash triage so engineers focus on feature work
Example Workflow
User asks: /crash-fixer --hours 12 --limit 3
- 1
Queries crash reporter API for crashes in the last 12 hours
- 2
Deduplicates results by fingerprint, skipping crashes already fixed
- 3
Analyzes each unique crash with AI to identify root cause
- 4
Generates a targeted code fix for each crash
- 5
Creates a branch, commits the fix, and opens a pull request on the target repo
Up to 3 pull requests opened on the target GitHub repo, each containing a fix for a deduplicated crash from the last 12 hours
Requirements
- GH_TOKEN environment variable (GitHub API token)
- CRASH_REPORTER_API_KEY environment variable
- CRASH_REPORTER_URL environment variable
- TARGET_REPO environment variable (GitHub repo in owner/name format)
Installation
clawhub install ryce/crash-fixerRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.