Claude Code Skills

Claude Code Skills

Reusable instruction packages that live under .claude/skills//SKILL.md, with optional reference files and scripts. Each skill is a folder containing a markdown file with YAML frontmatter (name, description, optional model, tools, hints) plus step-by-step instructions. Two types: capability uplift (pure prompt that teaches Claude to do something better) and encoded preference (sequential workflows tuned to the user). Skills can be invoked by slash command or natural language, run in the current session or delegate to sub-agents, and stack with each other.

Key facts

  • Stored at .claude/skills//SKILL.md within a project or globally in home directory
  • YAML frontmatter fields: name, description, disable-model-invocation, argument-hint, allowed-tools, model, hooks, agents
  • Two categories: capability uplift (prompt) and encoded preference (workflow)
  • Progressive context loading: only frontmatter loaded until skill selected, then skill.md, then references
  • Official Anthropic Skill Creator skill generates, tests, benchmarks, and trigger-tunes other skills
  • Can chain sub-agents, call scripts, and reuse reference markdown files
  • Skills can be installed via /plugins from the Anthropic marketplace
  • In the 10-hour course, skills are positioned as the mechanism for turning one-off successful Claude Code behaviours into repeatable capabilities that can be invoked later with natural language [src-016]
  • Skills sit alongside sub-agents, agent teams, browser automations, and context-management practices as the advanced layer after basic Claude Code workflow building [src-016]
  • Mornati’s MCP-vs-CLI comparison frames skills as a low-overhead middle layer for CLI tools: a GitHub CLI skill can steer agents toward reliable --json output while avoiding Native MCP’s always-on schema cost [src-041]
  • The same source warns that service-specific skill files should be invoked on demand; placing them in always-loaded project context turns them into recurring prompt overhead [src-041]

Source references

  • [src-004] Nate Herk cluster — Nate Herk — Claude Code cluster (21 videos)

– Videos referenced: zKBPwDpBfhs, RAZVk5NPNtE, mi4hcipESKQ, tXtCK66fPj8

  • [src-016] Nate Herk — “Build & Sell with Claude Code (10+ Hour Course)” (2026-03-12)
  • [src-041] Marco Mornati — “The Future of Agentic Tooling: MCP Servers vs. CLI A Data-Driven Comparison” (2026-04-27)

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 491 indexed pages and articles.

  1. Wiki concept Progressive Context Loading (Skills) The three-level loading pattern Claude Code uses to keep skills lightweight. Level 1: initial search only reads the YAML frontmatter (name + Related by skills
  2. Wiki concept Global vs Project Skills Claude Code skills can be scoped to a single project (live in .claude/skills/) or installed globally (live in ~/.claude/skills/) and available from Related by claude
  3. Insight AI Beyond POCs How enterprise AI moves beyond proofs of concept through ownership, governance, measurement, adoption, and production operating models Readers have engaged with this next