| name | hotfix |
| description | Emergency fix workflow — diagnose, apply minimal fix, test, publish, and post-mortem. Use when a live issue requires immediate action. |
| disable-model-invocation | true |
/hotfix — Emergency Fix Workflow
Delegate to: release-manager (with technical-director, relevant specialist for actual fix)
When to Use
- Live game broken for many users
- Critical security exploit discovered
- Data loss occurring
- Monetization broken
- Any issue that warrants interrupting normal sprint work
DO NOT Use For
- Minor bugs (use normal bug process)
- Features that didn't make the last release
- Polish items
Steps
1. Triage — Is This a Hotfix?
Confirm:
- Severity: Is this truly S0 or S1?
- Impact: How many users affected?
- Urgency: Can this wait for next scheduled release?
- Workaround: Is there a temporary workaround players can use?
If answer to "can this wait" is "yes" → not a hotfix. Use normal process.
2. Diagnose Fast
- Reproduce the issue (fast repro is critical)
- Identify root cause (not just symptom)
- Identify affected files/systems
- Delegate to the most relevant specialist for root-cause analysis
3. Design Minimal Fix
The fix should be:
- Minimal: Smallest possible change to fix the issue
- Targeted: Only touch files related to the bug
- Reversible: Easy to revert if the fix causes new issues
- Tested: Verify it fixes the issue without breaking neighbors
DO NOT:
- Bundle unrelated improvements
- Refactor around the bug
- Add new features
- Make "while we're here" changes
4. Test the Fix
- Reproduce the original issue → verify fix works
- Test neighboring functionality → verify no regression
- Test common user flows → verify no side effects
- If possible, test in a private place first
5. Expedited Review
- Quick code review (via lead-programmer or technical-director)
- Minimum checks: no hardcoded values, no client-trust, no new security holes
- If diff is < 20 lines, review should be < 10 minutes
6. Publish
- Publish directly to production place (skip staging if urgent)
- Use feature flag to enable fix if possible (instant rollback path)
- Communicate immediately via community channels
7. Monitor
- Watch metrics for 1-2 hours post-publish
- Verify fix is working for affected users
- Watch for new issues introduced by the fix
8. Post-Mortem (within 48 hours)
Every hotfix gets a post-mortem. Document:
- What broke: Clear description of the issue
- When it broke: When did this start affecting users?
- How it was detected: Player reports / monitoring / internal
- Root cause: Why did this happen?
- Fix: What was changed
- Prevention: What process change would prevent this in the future?
- Timeline: When it was detected → fixed → published
Save to production/incidents/hotfix-<date>-<summary>.md.
Output — Hotfix Report
# Hotfix: [Short description]
**Severity**: [S0/S1]
**Reported**: YYYY-MM-DD HH:MM
**Detected By**: [source]
**Fixed**: YYYY-MM-DD HH:MM
**Deployed**: YYYY-MM-DD HH:MM
**Duration of Impact**: X hours
## The Issue
[1-2 sentences on what broke]
## Root Cause
[1 paragraph on why]
## The Fix
- Files changed:
- [file 1]: [what changed]
- Lines changed: N
- Review: [who reviewed]
- Tests: [what was tested]
## Verification
- [How we verified the fix worked]
## Prevention
- [What to do to prevent this class of issue in the future]
## Action Items
- [ ] [Follow-up action]
- [ ] [Follow-up action]