Sentry → Auto-Debug → Open PR

{
  "webhooks": {
    "sentry": {
      "url": "/webhooks/sentry",
      "actions": [
        "analyze-error",
        "generate-fix",
        "create-pr",
        "notify-slack"
      ]
    }
  },
  "integrations": {
    "sentry": {
      "projectId": "your-project-id",
      "autoFix": true,
      "minOccurrences": 5
    },
    "github": {
      "repo": "owner/repo",
      "autoCreatePR": true,
      "assignReviewers": ["tech-lead"]
    }
  }
}

## On Sentry Error
- Only attempt a fix if the root cause is clear from the stack trace. If ambiguous, post a diagnostic summary to Slack instead of guessing.
- Never fix errors in hot paths (auth, payments, data writes) without flagging for human review — add a "needs-human-review" label to the PR.
- If the same error keeps recurring after a fix was deployed, escalate in Slack rather than opening another PR. Something deeper is wrong.
- PR descriptions should include: the Sentry issue link, a plain-English root cause explanation, and what you changed and why.
- Write regression tests that reproduce the exact error, not just generic happy-path tests.
- Ignore errors that are clearly client-side noise (e.g., browser extension interference, network timeouts from mobile).