How I use premortems with Claude and Codex

Short summary

Gary Klein's premortem technique—where you assume a plan has already failed and explain why—reveals sharper failure modes with LLMs than generic "what could go wrong?" prompts. The author built a Claude Code skill implementing this for coding decisions and product launches, producing specific failure stories and revised plans rather than generic risk lists. Works best with concrete plans and surfaces actionable corrections before resources are committed.

• •Premortem reframes planning: instead of 'is this good?' ask 'why did this already fail?'
• •Generates specific failure stories with LLMs—rollback boundaries, hidden assumptions—rather than generic lists
• •Published Claude Code skill with real examples: SQLite→Postgres migration, product launch plans

Generated with AI, which can make mistakes.