Arena Report

Per-Category Breakdown

For each of the 10 categories, include:

## <Category Name>

**Winner**: <adapter name> (score: X/5)

### Results by adapter

| Adapter | Score | Status | Notes |
|---------|-------|--------|-------|
| Claude | X/5 | PASS | ... |
| Codex | X/5 | PASS | ... |
| Cursor | X/5 | PASS | ... |
| Gemini | X/5 | PASS | ... |
| OpenCode | X/5 | PASS | ... |
| Pi | X/5 | PASS | ... |
| OpenClaw | X/5 | PASS | ... |

### Observations
<qualitative comparison of approaches taken by different adapters>

Overall Rankings

## Overall Rankings

| Rank | Adapter | Average Score | Strengths | Weaknesses |
|------|---------|---------------|-----------|------------|
| 1 | ... | X.X | ... | ... |
| 2 | ... | X.X | ... | ... |
| 3 | ... | X.X | ... | ... |
| 4 | ... | X.X | ... | ... |
| 5 | ... | X.X | ... | ... |
| 6 | ... | X.X | ... | ... |
| 7 | ... | X.X | ... | ... |

Scoring Rubric

Apply these criteria uniformly across all tasks and adapters:

• 5 - Excellent: Correct, complete, well-structured, follows best practices, handles edge cases
• 4 - Good: Correct and complete, minor style or structure issues
• 3 - Adequate: Mostly correct, some gaps or minor issues
• 2 - Poor: Partially correct, significant gaps or errors
• 1 - Failing: Incorrect, incomplete, or fundamentally flawed approach
• 0 - DNF: Agent failed to run, timed out, or did not produce output

Evaluation Dimensions

For each task result, consider:

1. Correctness -- Does the output satisfy the task requirements?
2. Completeness -- Are all parts of the task addressed?
3. Quality -- Code style, structure, best practices
4. Insight -- Does the agent show understanding beyond the literal task?
5. Efficiency -- Is the solution practical and performant?

Process

1. Collect all task results from runner agents' task comments
2. Parse the structured result format (Status, Output, Notes)
3. Score each result using the rubric and task-specific evaluation criteria
4. Generate the summary table
5. Write per-category breakdowns
6. Compute averages and produce rankings
7. Write the final report as a markdown document