What problem does it solve? Designing multi-agent systems by intuition leads to wrong orchestration patterns, invalid tool schemas, and undetected performance bottlenecks. This Skill replaces guesswork with three deterministic Python tools that score architecture patterns from requirements, generate provider-ready tool schemas, and analyze execution logs for cost, latency, and failure issues. ## Core Features & Use Cases - Architecture Planning: agent_planner.py scores and selects among single-agent, supervisor, swarm, hierarchical, and pipeline patterns from a requirements JSON, emitting agent role definitions, communication topology, a Mermaid diagram, and an implementation roadmap. - Tool Schema Generation: tool_schema_generator.py converts plain tool descriptions into validated OpenAI function-calling and Anthropic tool-use schemas with validation rules, error responses, rate limits, and usage examples. - Execution Log Evaluation: agent_evaluator.py computes success rates, latency percentiles, token costs, error patterns, and bottleneck analysis, then outputs prioritized optimization recommendations. - Use Case: Given requirements for a research automation platform, run the planner to get a supervisor architecture with a Mermaid diagram, generate validated Anthropic schemas for each agent's tools, then evaluate pilot run logs until zero critical issues remain. ## Quick Start Ask the agent to design a multi-agent architecture from your system requirements JSON and generate the corresponding validated tool schemas.
