If you're writing Claude Code hooks, this walks you through the entire lifecycle from design decisions to deployment. It's structured around progressive disclosure, loading detailed references only when you need them, which reportedly cuts token usage by 55%. The core content focuses on five specific pitfalls that break hooks in production: unquoted variables failing on filenames with spaces, missing input validation enabling path traversal, synchronous operations blocking Claude for 30+ seconds, using exit 1 instead of exit 2 so PreToolUse hooks don't actually block, and assuming dependencies exist. Good for both writing new hooks and debugging why existing ones fail intermittently.
npx -y skills add secondsky/claude-skills --skill claude-hook-writer --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