Claude Code Skills: How to Create, Install, and Verify Them

Claude Code skills are reusable folders centered on a SKILL.md file containing YAML metadata and instructions, designed to provide procedures, references, scripts, or checks that load fully only when invoked while their names and descriptions consume a limited per-turn context budget. The walkthrough explains how to create a project-level ui-check skill that identifies changed front-end files and uses Kane CLI to verify visible behavior in a real browser, including branching on command exit codes to distinguish product failures from environment issues. It covers skill locations, naming precedence, differences between skills and CLAUDE.md files, hooks, MCP servers, commands, and plugins, as well as frontmatter validation, marketplace installation, tool permissions, and optional fields controlling invocation, visibility, context, paths, and model settings. The text emphasizes keeping descriptions concise and trigger-focused, limiting skill-body size, monitoring context use with Claude Code diagnostic commands, and evaluating skills both for correct triggering and measurable output improvements through baseline comparisons and automated evaluations. Its dark-mode example demonstrates why UI checks should validate rendered results rather than superficial signals such as a changed button label, with browser evidence helping agents identify and correct the underlying CSS or DOM issue.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.