If you have been using Claude Code by typing the same instructions over and over, you are leaving enormous productivity on the table.
Slash commands are Claude Code's most underrated feature. They let you create reusable, parameterized workflows that run instantly - no copy-pasting, no re-explaining context, no waiting.
I run an e-commerce business with a team of 90+ AI agents, all coordinated through Claude Code. Slash commands are the glue that holds the whole system together.
Here is exactly how I use them, and how you can set them up today.
What Are Slash Commands?
Slash commands are custom instructions stored as Markdown files in .claude/commands/. When you type /your-command-name in Claude Code, it loads and executes that file instantly.
Think of them as macros for your AI workflow - but smarter, because they can include context, examples, constraints, and even call other commands.
Structure:
.claude/
commands/
review.md
deploy-check.md
write-email.md
You type /review and Claude Code loads review.md and executes it with full context of your project.
Why This Changes Everything
Before slash commands, every session I had to re-explain:
- My brand voice
- My tech stack
- What "done" means for each task
- Which files to avoid touching
With slash commands, all of that context is pre-loaded. I type /campaign-review and my entire brand guidelines, past performance data reference, and output format are baked in automatically.
The time savings compound. If each task saves 5 minutes of re-explaining, and you run 20 tasks per day, that is 100 minutes saved daily. Over a month: 50 hours.
The 5 Slash Commands I Use Every Day
1. /review - Code and Content Quality Gate
This is my most-used command. It runs a 7-lens review on any output before it goes live.
# Review Command
You are a quality reviewer. Run these 7 checks on the content or code I provide:
1. **Technical accuracy** - are all facts verifiable?
2. **Brand voice** - matches our tone (direct, no fluff, data-driven)?
3. **Security** - no exposed credentials or vulnerable patterns?
4. **Performance** - any obvious bottlenecks?
5. **Completeness** - does it fully solve the stated problem?
6. **Edge cases** - what could break?
7. **Next steps** - what should be done after this?
Output: GO / NO-GO with specific line-by-line notes.
Want the full 7-lens review template? Get it at mynextools.com - free with account signup.
2. /router - Task Routing to the Right Agent
I have specialized agents for SEO, email, ads, analytics, and more. The /router command analyzes a free-text request and assigns it to the correct agent automatically.
# Router Command
Analyze the incoming request and determine which team member should handle it.
Teams available:
- /analytics - data analysis, reporting, metrics
- /content - writing, editing, SEO articles
- /ads - Meta/Google campaign management
- /email - email sequences, Klaviyo automation
- /operations - task tracking, system maintenance
Output: agent name + reasoning + handoff message ready to paste.
This alone saved me from routing errors that were costing hours every week.
3. /morning-briefing - Daily Status in 60 Seconds
Every morning this command pulls together everything I need to know: open tasks, yesterday's performance flags, and today's priorities.
# Morning Briefing
Read the following files and generate a morning briefing:
1. משימות-פתוחות.md - open tasks flagged as urgent
2. דוחות/מטא/ - latest Meta ads report
3. קריאייטיב/nex-ig-queue.json - content queue status
Output format:
- 🔴 Urgent (must do today)
- 🟡 Important (should do today)
- 🟢 Queue (do when possible)
- 📊 Key metric from yesterday
4. /session-close - End-of-Session Handoff
This is the command that makes multi-session work actually work. It saves everything I need to pick up exactly where I left off.
# Session Close
Before ending this session:
1. List all changes made (files created/edited)
2. List all decisions made with reasoning
3. List all open items and blockers
4. Write a "next session prompt" that will resume this work
5. Save to פרומפט-{topic}-{date}.md
6. Update MEMORY.md with any new learnings
Format: structured, scannable, nothing implicit.
5. /fact-check - Zero Fabrication Policy
This one is critical for content that goes public. It flags any claim that is not sourced.
# Fact Check
Review the content for any factual claims:
- Statistics (must have year + source)
- Product comparisons (must be current)
- "Research shows" statements (must cite specific research)
- Any number that could mislead
Flag: [UNVERIFIED] next to any claim without a source.
Output: clean version + list of items needing verification.
How to Create Your First Slash Command
Step 1: Create the directory if it does not exist:
mkdir -p .claude/commands
Step 2: Create a Markdown file:
touch .claude/commands/my-command.md
Step 3: Write your instruction in plain language. Include:
- What problem this solves
- What input you will provide
- What output format you expect
- Any constraints or rules
Step 4: Test it by typing /my-command in Claude Code.
That is it. No configuration files, no YAML, no build step.
Pro tip: Keep a
commands/README.mdlisting all your commands with one-line descriptions. As your library grows, this becomes essential.
Advanced Patterns
Chaining Commands
Commands can reference other commands. My /weekly-report command calls /analytics then /content then /ads in sequence.
# Weekly Report
Run the following sequence:
1. /analytics - pull 7-day performance data
2. /ads - audit current campaign status
3. /content - check queue levels
Then synthesize into a single executive summary with:
- Top 3 wins
- Top 3 problems
- Recommended actions for next week
Parameterized Commands
You can pass context when you invoke a command. Type /review followed by the content you want reviewed, and the command receives it as input.
Global vs Project Commands
-
Global:
~/.claude/commands/- available in every project -
Project:
.claude/commands/- scoped to one project
I keep general utilities global (review, fact-check, router) and project-specific workflows local.
The System I Built
My full command library has 40+ slash commands organized by department:
-
/ראוטר- main router -
/מנכל- CEO-level strategic decisions -
/קמפיינים- campaign performance -
/מפעל-תוכן- content factory workflow -
/שומר- end-of-session save
Each command is under 50 lines. Together they automate decisions that used to require manual coordination.
The full command library template is available at mynextools.com. Includes 15 production-ready commands you can adapt immediately.
Getting Started: The 3-Command Starter Pack
If you want to start small, create these three commands first:
-
/review- quality gate for everything you produce -
/session-close- ensures nothing is lost between sessions -
/morning-briefing- daily context loading in 60 seconds
These three alone will change how you work with Claude Code.
Conclusion
Slash commands are not a power-user trick. They are a fundamental shift in how you interface with Claude Code.
The difference between typing instructions every session versus loading a pre-built command is the difference between using a calculator and building a spreadsheet. Both do math, but one scales.
Start with one command. Get it working. Then build the next one. In two weeks you will have a personal automation library that would have taken a team months to build manually.
The documentation is at docs.anthropic.com if you want to go deeper. But honestly, the best way to learn is to create one command right now and use it today.
Built this with Claude Code at mynextools.com - tools for founders who move fast.
Tags: claude-code, productivity, ai-tools, developer-tools, automation
Top comments (0)