I built a Claude Code skill for structured decision making
The article introduces a 'Six Hats' skill for AI agents that facilitates structured decision-making using Edward de Bono's six thinking hats framework. It guides users through sequential rounds of analysis—facts, emotions, benefits, risks, alternatives, and final synthesis—to produce actionable recommendations. The skill is designed for evaluating decisions in technology, career planning, and strategy with concrete prompts and outputs.
- ▪The Six Hats skill uses AI to simulate six distinct thinking modes: White Hat (facts), Red Hat (emotions), Yellow Hat (benefits), Black Hat (risks), Green Hat (alternatives), and Blue Hat (synthesis).
- ▪It is intended for use in decision-making scenarios such as technology migrations, career changes, or product strategy evaluations.
- ▪Example applications include debating a move from frontend tooling to an AI company and assessing React's viability for frontend development in 2026.
- ▪The skill produces detailed, structured debates and saves outputs to specified directories for review.
- ▪Installation requires cloning the GitHub repository into the appropriate skills directory for the AI agent platform being used.
Opening excerpt (first ~120 words) tap to expand
Six Hats Skill A structured decision-debate skill for running Edward de Bono-style six hats sessions with an AI agent. It walks a topic through facts, intuition, upside, risk, alternatives, and final moderation so you get a practical recommendation instead of a loose brainstorm. What It Does The skill runs three sequential rounds using these roles: White Hat: facts, known information, and unknowns Red Hat: intuition, emotional reactions, and gut checks Yellow Hat: benefits, upside, and reasons the idea could work Black Hat: risks, objections, failure modes, and downsides Green Hat: alternatives, reframes, and creative options Blue Hat: moderation, synthesis, final recommendation, agreements, tensions, and next steps The hats run in order so each perspective can build on the previous ones.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.