Bug Triage Agent

Automated GitHub Issue Triage That Routes Bugs Before They Become Fires

Every engineering team has the same problem: bugs get filed, nobody owns them, and three weeks later a P0 turns into a customer escalation. The Bug Triage Agent solves this at the source — automatically classifying, prioritizing, and routing every GitHub issue the moment it's filed.

What It Does

This workflow runs as a GitHub webhook handler, triggering instantly when a new issue is opened or reopened on your repository. Here's what happens within seconds of a bug being filed:

1. Issue Analysis: The workflow parses the full issue — title, body, reporter information, existing labels, and repository context. It checks for duplicate issues by comparing against recently closed and open issues to avoid duplicate triage.

2. AI Classification: The issue is analyzed by Claude to determine:

• Type: bug, feature request, question, or duplicate
• Severity: P0 (critical, no workaround), P1 (major with workaround), P2 (standard bug), P3 (minor/cosmetic)
• Affected area: auth, payments, onboarding, performance, UI, API, or general
• Suggested assignee: based on your configured ownership map

3. Automated Actions: Labels are applied, an assignee is suggested, and a triage comment is posted directly on the GitHub issue explaining the classification and next steps.

4. Critical Alerts: P0 and P1 issues trigger an immediate Telegram alert with the issue title, severity, link, and assigned owner — no waiting for someone to notice it in the issue tracker.

5. Enterprise Customer Flags: If the reporter's company matches your configured enterprise customer list, the issue is flagged with elevated priority regardless of severity classification.

6. SLA Tracking: Response time expectations are logged based on severity: P0 = 2 hours, P1 = 24 hours, P2 = 72 hours, P3 = 1 week.

Who This Is For

Engineering teams of any size that have a public or semi-public GitHub repository and deal with a consistent volume of bug reports. Most impactful when you have more than 20 issues per week — enough that manual triage becomes a real time sink.

What's Included

• bug-triage-agent.yml — complete workflow definition with webhook trigger, AI triage steps, GitHub actions, and Telegram alerting
• Team routing configuration (ownership map by area)
• Enterprise customer list configuration
• SLA thresholds per severity level
• Duplicate detection logic

Setup Steps

1. Install the workflow — copy bug-triage-agent.yml into your OpenClaw workspace
2. Set environment variables — add GITHUB_TOKEN and GITHUB_WEBHOOK_SECRET to your OpenClaw environment
3. Configure team routing — edit the team_routing section to map code areas to GitHub handles
4. Add your enterprise customer list — list any customer names that should trigger elevated priority
5. Register the GitHub webhook — point it to your OpenClaw webhook endpoint (/webhooks/github-issues) with issues events enabled
6. Test it — open a test issue; within seconds it should be classified, labeled, and a triage comment should appear

Why This Beats Manual Triage

Without automated triage, someone on your team has to read every new issue, decide its priority, figure out who should own it, and post a response. For teams with 20+ issues/week, that's 2–4 hours of triage work per week — often done inconsistently.

With this workflow, triage happens in seconds, consistently, 24/7. P0 issues get flagged before your on-call person might notice them. Enterprise customer bugs get elevated automatically. Your engineers see a labeled, assigned, triaged backlog instead of an undifferentiated pile.

Result: faster response to critical bugs, cleaner backlog, and no more issues falling through the cracks.