Protocol Thunderdome
"Two devs enter. One codebase."
AI-powered scrum master for GitHub team collaboration. Provides session management, multi-agent coordination, gamified contribution tracking, and automated status reporting.
Commands
| Command | Description |
|---|---|
/thunderdome | Full status report |
/thunderdome status | Same as above |
/thunderdome debrief | End-of-session checklist |
/thunderdome scores | Gamification leaderboard |
/thunderdome agents | Multi-agent coordination status |
Status Report Workflow
When invoked, generate a comprehensive status report:
1. Git State
# Fetch latest
git fetch origin --quiet
# Check uncommitted files
git status --porcelain
Report:
- CRITICAL: X uncommitted files (list first 5)
- WARNING: X unpushed commits
- OK: Working directory clean
- INFO: Current branch
2. GitHub State
# Recent commits
gh api repos/{owner}/{repo}/commits --jq '.[:5] | .[] | "\(.sha[0:7]) \(.author.login): \(.commit.message | split("\n")[0])"'
# Open PRs
gh api repos/{owner}/{repo}/pulls --jq '.[] | "PR #\(.number): \(.title)"'
# Open issues (excluding PRs)
gh api repos/{owner}/{repo}/issues --jq '.[] | select(.pull_request == null) | "#\(.number) \(.title)"'
# Branches
gh api repos/{owner}/{repo}/branches --jq '.[].name'
Report:
- X open PRs awaiting review (list them)
- X issues in backlog
- Recent commits by contributor
- Branch inventory
3. Test Status
Detect and run project tests:
package.json→npm testMakefile→make testswiftproject →swift testpytest→pytest
Report:
- Tests: X passed, Y failed
- CRITICAL if any failures
4. Contributor Stats
For each contributor, count:
- Commits today
- Commits this week
- Calculate gamification score
5. Summary
End with assessment:
- ALL CLEAR - Ready for battle
- WARNINGS - List items needing attention
- CRITICAL - Blockers that must be resolved
Debrief Protocol
Before ending a session, verify:
1. Tests Pass (MANDATORY)
Run full test suite. DO NOT proceed if tests fail.
- Fix failures, OR
- Document as known issue in session log
2. No Uncommitted Changes
git status --porcelain | wc -l
If changes exist:
git add -A
git commit -m "Session: <brief description>"
3. Changes Pushed
git log origin/main..HEAD --oneline
Push any unpushed commits:
git push origin main
4. Documentation Current
Check if key docs need updates:
- README.md
- CLAUDE.md (if exists)
- Any docs describing changed behavior
5. Issues Synced
- Close completed issues
- Create issues for discovered bugs
- Update labels (in-progress, done)
6. Final Checklist
Output verification:
- All tests passing
- All changes committed
- All commits pushed
- Documentation updated
- Issues synced
Only report SAFE TO CLOSE when all items checked.
Gamification System
Track contributions with competitive scoring:
Scoring Rules
| Action | Points |
|---|---|
| Commit with tests | +50 |
| Small commit (<50 lines) | +10 |
| PR merged | +100 |
| Bug fix | +25 |
| Review submitted | +15 |
| Breaking CI | -100 |
| Untested code dump (>300 lines) | -75 |
| Lazy commit message (<10 chars) | -15 |
| Force push to main | -50 |
Titles (Score-Based)
| Score | Title |
|---|---|
| 0-99 | Keyboard Polisher |
| 100-299 | Bug Whisperer |
| 300-499 | Commit Crusader |
| 500-999 | Merge Master |
| 1000-1999 | Pull Request Paladin |
| 2000-3999 | Refactor Ronin |
| 4000-7499 | Test Titan |
| 7500-14999 | Pipeline Pharaoh |
| 15000+ | Code Demigod |
Shame Titles
Awarded for consistent bad behavior:
- Build Breaker - 3+ CI failures in a week
- YOLO Developer - No tests in 5+ consecutive commits
- Benchwarmer - Lowest weekly score among active contributors
Multi-Agent Coordination
When multiple Claude Code instances work on the same repo:
Intent Broadcasting
Before starting work:
- Check if other agents are registered
- Declare what you're working on
- Check for file conflicts
Safe Parallel Zones
These can be edited simultaneously WITHOUT coordination:
- Different views/components
- Backend vs frontend directories
- Tests vs implementation (different files)
- Separate documentation files
Conflict Resolution Priority
- First registered wins - Agent who declared first has priority
- Critical path wins - Bug fixes > features > refactoring
- Ask the human - When priority unclear
Agent Coordination File
If .thunderdome/agents.json exists, use it:
{
"agents": [
{
"id": "session-abc123",
"task": "Implementing feature X",
"files": ["src/feature.ts", "src/feature.test.ts"],
"started": "2026-01-28T10:00:00Z"
}
]
}
Register your task, check for conflicts, deregister when done.
PR/Code Review Workflow
AI manages code review when asked:
Creating PRs
gh pr create --title "Feature: description" --body "## Summary
- Change 1
- Change 2
## Test Plan
- Verified X
- Tested Y"
Reviewing PRs
# View diff
gh pr diff <number>
# Approve
gh pr review <number> --approve --body "LGTM - tests pass, code clean"
# Request changes
gh pr review <number> --request-changes --body "Please fix: ..."
Review Checklist
- Code compiles/builds
- Tests pass
- No secrets committed
- Commit messages descriptive
- Related issue linked
- No unnecessary changes
Merging
# Merge with commit (preserves history)
gh pr merge <number> --merge --delete-branch
# Or squash (cleaner history)
gh pr merge <number> --squash --delete-branch
Configuration
Create .thunderdome/config.json for project-specific settings:
{
"contributors": ["username1", "username2"],
"gamification": true,
"testCommand": "npm test",
"criticalFiles": ["src/models/*", "src/services/*"],
"coordinationEnabled": true
}
Philosophy
Thunderdome treats collaborative development as a competitive-cooperative sport:
- Compete on contribution quality (tests, atomic commits, clean code)
- Cooperate on shared goals (shipping features, maintaining quality)
- Coordinate through visibility (everyone sees the same state)
The gamification incentivizes good practices:
- Tests with code (+50) beats untested dumps (-75)
- Small atomic commits (+10) beat monolithic changes
- Passing CI beats breaking the build (-100)
May your commits be atomic and your tests green.
