Home / CCA-F / Claude Code / Slash Commands, Skills, and Subagents — Detail & Examples
DWG 3.3 — DETAIL & EXAMPLES

Slash Commands, Skills, and Subagents — Detail & Examples

Part of Claude Code · overview: domain-3.html
In this page

A worked example choosing between a slash command, a skill, and a subagent for the same team need, and a quick test for delegating vs. handling work inline.

3.3aSlash command vs. skill vs. subagent

WORKED EXAMPLE · PICKING THE RIGHT MECHANISM
TASKA team wants Claude Code to reliably run their release process, which involves several specific steps done the same way every time.
  1. If it's one specific, short, repeatable instruction ("run the release checklist") — a slash command is enough: it's a shortcut for typing the same prompt every time.
  2. If the release process involves several files of procedural knowledge, examples of past releases, and maybe a helper script — that's a skill: more than a single prompt, packaged for reuse across many tasks.
  3. If, mid-task, Claude needs to do a large chunk of self-contained work (e.g. auditing every changelog entry since the last release) that would otherwise fill up the main session's context — delegate that specific chunk to a subagent and bring back only its summary.
Distinguishing question:Is this a shortcut for a prompt (slash command), packaged reusable knowledge (skill), or self-contained work that should be context-isolated (subagent)?

3.3bDelegate or handle inline?

A quick test
  • Inline: the work is short, or depends heavily on context already in the main conversation (e.g. "now rename that variable we just discussed").
  • Subagent: the work is long, self-contained, and would otherwise flood the main session with detail unrelated to the main task (e.g. "scan the entire codebase for any use of a deprecated API").
← 3.2 CLAUDE.md Configuration Claude Code Overview 3.4 Hooks, SDK & CI/CD →
Claude Certified Architect — Study NotesDWG 3.3