Forces you to argue against your first instinct before making architectural decisions or choosing between approaches. It's designed to interrupt the pattern where LLMs lock onto an answer early and rationalize backward. When you're about to recommend a framework, implementation strategy, or technical direction, it makes you steel-man the opposition, identify non-obvious failure modes, and state your assumptions explicitly. Skip it for trivial choices or when you're just executing an already-made decision. Works well paired with reasoning-verifier afterward to catch any remaining gaps. Honest take: this adds friction, but the kind that prevents you from confidently walking into predictable traps.
npx -y skills add majesticlabs-dev/majestic-marketplace --skill devils-advocate --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
shadcn/improve
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills