This is a opinionated ADR workflow built for AI coding agents, not just humans documenting decisions after the fact. It enforces a four phase process: scan the codebase first, then Socratic questioning to capture intent, then drafting with explicit implementation plans that tell agents which files to touch and how to verify correctness, then validation against an agent readiness checklist. The core insight is smart: if Claude is going to implement your architectural decisions, the ADR needs constraints that are measurable, not vibes, and enough specificity that no follow up questions are needed. It even tells agents when to proactively stop and propose an ADR before introducing new dependencies or patterns.
npx -y skills add skillrecordings/adr-skill --skill adr-skill --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