Skip to content
Home › Claude Code Configuration & Workflows

Skills And Slash Commands

Claude Code Configuration & Workflows

Definition

Reusable, on-demand instruction bundles stored under .claude/ that surface as /slash-commands, whose body is loaded into context only when invoked — keeping the base prompt small.

Key points

  • Live in .claude/commands/ or .claude/skills/SKILL.md; each becomes a /slash-command.
  • The command/skill body is loaded only when invoked, not at every turn — this is the Minimal footprint principle applied to memory: metadata is cheap, the full instructions are lazy-loaded.
  • Skills can also be attached to subagents via the skills field in an AgentDefinition / filesystem agent frontmatter.
  • Sits alongside the always-loaded CLAUDE.md Hierarchy (concatenated every turn) — slash commands are the opt-in counterpart.

Why it matters for the exam

  • D3 (20% of scored content) tests knowing that slash-command bodies are lazy-loaded (only on invocation), which is the correct answer when a scenario asks how to add capability without bloating every prompt.

Common gotchas

  • The body is not in context until the command is invoked — don't assume a skill's instructions are always active.
  • Distinguish from CLAUDE.md, which loads its full content every session.

See also

Sources

Referenced by