claude-skills-toolkit

Implement skills using AskUserQuestion tool for interactive user input and decisions. Use when authoring skills that need to gather options, handle free-text answers, process multi-select, or build multi-step workflows. Covers tool mechanics, constraints, wiring into skills, response handling, and dynamic question patterns.

Create NEW Claude Code skills from scratch following best practices. Use when building new skills, interviewing for requirements, applying templates, organizing frontmatter and body content, or converting slash commands to skills. Guides skill structure, naming, descriptions, progressive disclosure, reference organization, and tool scoping.

Create, validate, and refine Claude Code subagents for reliable delegation. Use when building new subagents, validating existing ones, improving quality, scoping tool access, configuring permission modes, or setting up hook validation. For personal, team, or production environments.

Test and benchmark Claude Code skills empirically using evaluation-driven development. Use when validating a skill's effectiveness, comparing skill vs. baseline performance, running benchmarks with timing/token metrics, and iterating on skill improvements. Two modes: Quick Workflow (with_skill only, no baseline, fast pass/fail) or Full Pipeline (7-phase with baseline comparison, timing, aggregation).

Create, convert, validate, and publish Claude Code plugins with Agent Skills, hooks, agents, and servers. Use when building plugins from scratch, converting projects to plugins, improving plugin structure, or publishing to marketplace. Includes automated scanning, manifest generation, marketplace.json creation, and validation guidance. Component-specific work delegates to hook-creator, subagent-creator, and skill-creator skills.

Improve and validate existing Claude Code skills for clarity, efficiency, and production readiness. Use when: refining skills, improving skill structure, validating against best practices, reducing token usage, consolidating references, checking production readiness, or applying the 80% rule. Takes imperfect existing skills and elevates them to quality standards.

Create, validate, and refine Claude Code plugin hooks for automating workflows. Use when building new hooks from scratch, validating existing hooks against best practices, or improving hook quality. Supports command hooks (shell scripts), prompt hooks (LLM decisions), event matching, decision schemas, and production safety validation. Claude auto-activates when you ask to build a hook, check hook reliability, improve hook configurations, or validate hooks for production.