This is a structured approach to writing tickets that AI agents can execute without asking follow-up questions. The core insight is that tickets are prompts, so they need explicit file references, pattern examples, verification commands, and constraints instead of vague descriptions that assume human context. It extends INVEST criteria with a seventh "Claude-Ready" dimension and provides templates for features, bugs, and tech debt that include sections like "Relevant Files," "Pattern Reference," and actual bash commands for verification. You'd use this when breaking down work for Claude Code or other AI agents, especially if you're tired of agents asking "which auth module?" or "what conventions?" The templates are verbose by design because specificity is the point.
npx skills add https://github.com/alinaqi/claude-bootstrap --skill ticket-craft