brainstorming_skill

Overview

This skill designs exploration workflows by spawning parallel agents to turn rough ideas into validated designs. It auto-detects project complexity, selects appropriate agents, runs a seven-phase brainstorming process, and returns trade-offs and test strategies for the top options. Use it to compare alternatives, surface feasible approaches, and produce presentation-ready design summaries.

How this skill works

The skill scans project signals (README, CI, Docker, infra) to classify a project tier and constrain solutions. It dynamically selects domain agents (system design, frontend, backend, security, testability) and runs a seven-phase workflow: topic analysis, memory retrieval, divergent ideation, feasibility fast-checks, evaluation with devil's-advocate, synthesis, and a final design presentation. Tasks are created for each phase to provide progress visibility, and an optional Agent Teams mode enables real-time debate between agents.

When to use it

• When you need many diverse ideas and structured comparison
• When exploring multiple architectural or feature alternatives
• When you want designs constrained by detected project maturity
• When you need designs accompanied by test strategies and feasibility checks
• When you want parallel agent research to speed discovery

Best practices

• Run the auto-detection step first and accept or override the tier with awareness of trade-offs
• Prefer Agent Teams for open exploration and deep evaluation; use Task Tool for quick ideation or constrained work
• Always include a testability agent to score designs and propose test plans
• Generate a broad idea pool (10+) before filtering — divergent-first yields better trade-offs
• Create and track tasks for each phase so outputs are auditable and reviewable

Example use cases

• Brainstorm API and system architecture options with backend, workflow, and testability agents
• Explore UI/UX approaches for a dashboard while including frontend and accessibility perspectives
• Compare RAG pipeline designs with llm-integrator and data-pipeline agents and receive test strategies
• Evaluate caching or scaling approaches and get feasibility checks and trade-off tables
• Conduct a quick ideation sprint to surface 10+ ideas, then synthesize the top 3 with test plans