What problem does it solve? Creating Claude Code extensions requires knowing many scattered rules: SKILL.md frontmatter formats, sub-agent file conventions, plugin manifests, hook events, settings hierarchy, and sandboxing options. This Skill consolidates the official documentation and best practices into one authoring reference so extensions are built correctly the first time. ## Core Features & Use Cases - Extension Authoring Reference: Covers Skills (progressive disclosure, frontmatter rules), sub-agents (file format, nesting limits), plugins (plugin.json manifest), slash commands (parameters, $ARGUMENTS), hooks (PreToolUse, SessionStart), memory (CLAUDE.md), settings (6-level hierarchy), sandboxing, and headless mode. - Advanced Agent Patterns: Documents two-agent initializer/executor patterns, orchestrator-worker architectures, context engineering principles, and tool design best practices drawn from Anthropic engineering posts. - Quality Gates: Ships a 200+ point best-practices checklist plus rationalization, red-flag, and verification sections to catch common authoring mistakes like invalid YAML, missing hook scripts, or oversized CLAUDE.md files. - Use Case: When asked to create a new project Skill or configure a PreToolUse hook, load this Skill to get the exact frontmatter fields, directory layout, and validation steps. ## Quick Start Ask the AI to create a new Claude Code Skill or sub-agent for your project and it will apply the official formats and checklists from this kit.
