mcp-code-execution-results-comparison-analyzer

Compare AI agent performance by analyzing experimental results using automated extraction of metrics, logs, and workspace outputs. This Skill processes zipped experiment data to generate side-by-side comparisons of duration, cost, token usage, and output quality across code-execution versus direct-MCP approaches, producing a comprehensive report with actionable recommendations.

Extract duration, cost, token usage, efficiency ratios, success/failure rates, and workspace outputs from experiment result archives. The Skill aggregates these metrics across successful runs, catalogs output files for qualitative assessment, and identifies failed runs with clear summaries to help you understand where each approach excels.

Yes, this Skill adapts to varying zip structures and automatically identifies failed runs. It handles different experiment layouts and data organizations, extracts metrics and logs regardless of archive structure, and produces consistent comparison reports with recommendations tailored to your specific experimental setup.

Process your experiment results archive through this Skill to compare performance, cost, and token usage across both approaches. The automated analysis identifies efficiency ratios, highlights success rates, and generates recommendations on when to use each approach based on your specific task requirements and constraints.

The Skill generates a comprehensive comparison report with performance metrics, cost analysis, token usage data, failure summaries, output quality excerpts, and actionable recommendations. Results are saved to ./workspace, providing a complete audit trail for optimizing AI agent workflows.

No, this Skill automates data extraction entirely. It reads zipped experiment results, parses metrics and logs automatically, identifies and summarizes failures, catalogs workspace outputs, and aggregates data across runs—eliminating manual analysis and providing structured comparison insights immediately.