Building Custom Skills and Slash Commands in Claude Code

If you find yourself typing the same long instructions into Claude Code over and over (like "review this PR against this checklist..."), turn it into a reusable skill or slash command — write it once, reuse it forever.

What Is a Slash Command?

A slash command is a shortcut — typing /deploy triggers a pre-written set of instructions. These live as simple markdown files inside a project's .claude/commands/ folder.

<!-- .claude/commands/review.md -->
Review the current diff and check the following:
1. Any security issues (SQL injection, XSS)?
2. Does naming follow the project's conventions?
3. Was any unnecessary complexity added?

Give the findings as a bullet list.

Now typing just /review in the terminal runs this entire checklist automatically.

What Is a Skill?

A skill is a bit more structured — a folder containing a SKILL.md file with metadata (name, description) and detailed instructions. Skills load automatically when a task matches them — like a "database-migration-checker" skill that only activates when you're talking about migrations.

---
name: migration-safety-check
description: Reviews new database migrations for production safety (locking, large table alters, backward compatibility).
---

Whenever you see a new migration file, verify:
- Is it adding a NOT NULL column to a large table without a default?
- Is the rollback (down) method correctly defined?
- Is it dropping any existing column that might still be used elsewhere?

Skill vs Slash Command — When to Use Which?

• Slash command — a small, direct "do this one thing right now" task.
• Skill — larger, contextual knowledge that should apply automatically when relevant, without the user explicitly invoking it.

Key Takeaways

• Turning repeated prompts into slash commands saves time.
• Skills package complex, reusable domain knowledge that can trigger automatically.
• Both can be shared with a team (by committing them to the repo) so everyone uses the same workflow.