Agent Skills Fundamentals

Reference guide to how Agent Skills actually work under the hood in Claude Code - the SKILL.md structure and frontmatter fields (name, description, allowed-tools, model), how descriptions drive matching and confirmation-before-load, the Enterprise > Personal > Project > Plugins priority order for name conflicts, progressive disclosure via scripts/references/assets, how skills differ from CLAUDE.md/slash-commands/hooks/subagents/MCP, the gotcha that subagents don't inherit skills unless listed in their frontmatter, and a troubleshooting checklist for skills that won't trigger, won't load, or fail at runtime. Use whenever the user is authoring, debugging, sharing, or reason…