Competitive coding tournament - N agents compete, judge picks winner. Usage: /tournament [task] [competitors:5]
Orchestrates competitive coding tournament with multiple AI agents competing on a task
/plugin marketplace add ANcpLua/ancplua-claude-plugins/plugin install workflow-tools@ancplua-claude-pluginsTask: $1 Competitors: $2 (default: 5)
<CRITICAL_EXECUTION_REQUIREMENT> THIS IS A COMPETITION. YOU ORCHESTRATE, AGENTS COMPETE.
ā ļø YOU ARE THE JUDGE, NOT A COMPETITOR:
ā TOURNAMENT RULES:
YOUR NEXT MESSAGE: Launch N Task tool calls (one per competitor). NOTHING ELSE. </CRITICAL_EXECUTION_REQUIREMENT>
For EACH competitor, launch a Task with:
subagent_type: feature-dev:code-architect
model: opus
description: "Tournament competitor N"
prompt: |
š TOURNAMENT COMPETITION š
You are a competitor in a coding tournament.
Other competitors are working on the SAME task.
Only the BEST solution wins.
TASK: [insert $1 here]
RULES:
- Write the BEST, most elegant solution
- Be AGGRESSIVE - don't just fix, OPTIMIZE
- Show your work: explain WHY your solution is superior
- Code must compile and pass tests
SCORING CRITERIA:
1. Correctness (40%) - Does it work?
2. Elegance (25%) - Is it clean and readable?
3. Performance (20%) - Is it efficient?
4. Completeness (15%) - All edge cases handled?
At the end, explain why YOUR solution should win.
Output:
- All code changes (full files or diffs)
- Explanation of approach
- Why this solution is BEST
After ALL competitors complete, launch ONE judge:
subagent_type: feature-dev:code-reviewer
model: opus
description: "Tournament judge"
prompt: |
š TOURNAMENT JUDGE š
You are judging N solutions for the task.
SCORING (100 points total):
- Correctness: 40 pts (compiles, tests pass, no bugs)
- Elegance: 25 pts (clean, readable, idiomatic)
- Performance: 20 pts (efficient, no waste)
- Completeness: 15 pts (edge cases, error handling)
FOR EACH COMPETITOR:
1. Score each criterion
2. Total score
3. Specific praise
4. Specific criticism
FINAL RANKING:
1. š„ Winner: [name] - [score]/100
2. š„ Second: [name] - [score]/100
3. š„ Third: [name] - [score]/100
WINNER'S SOLUTION:
[Show the winning code that should be applied]
Launch ONE implementer:
subagent_type: feature-dev:code-architect
description: "Implement winner solution"
prompt: |
IMPLEMENT the winning solution from the tournament.
Apply all the winning code changes.
Ensure tests pass.
Format code properly.
Output: Files changed with verification
dotnet build 2>&1 || npm run build 2>&1 || make build 2>&1
dotnet test 2>&1 || npm test 2>&1 || make test 2>&1
After completion, output:
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā š TOURNAMENT RESULTS š ā
ā āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā£
ā Task: $1 ā
ā Competitors: N ā
ā āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā£
ā FINAL STANDINGS ā
ā š„ 1st: [Competitor X] - [score]/100 ā
ā š„ 2nd: [Competitor Y] - [score]/100 ā
ā š„ 3rd: [Competitor Z] - [score]/100 ā
ā āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā£
ā Winning Solution Applied: ā
ā
ā Build: PASS/FAIL ā
ā Tests: PASS/FAIL ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā