Home / Shortcuts
◆

The Shortcuts.
Small moves, big payoff.

Bonus reference
Useful across all 3 stages
i

The real shortcuts aren't keyboard tricks. They're the prompt patterns and Skills/GPTs/Gems you build yourself. The keyboard stuff is nice; the personal pack is the moat.

A · Keyboard

Claude.ai keyboard shortcuts

These are the ones I actually use daily. Cmd+? in any Claude chat shows the full current list (it updates regularly).

2 min read
Cmd+?
Show all shortcuts for your platform. The cheat sheet inside the cheat sheet.
Cmd+K
New chat instantly. Don't pollute a thread with a different topic.
↑
Edit your last message (input must be empty first). The single biggest token-saver on long threads.
Cmd+.
Stop generation mid-response. When you can tell from the first paragraph it's off, stop and re-prompt.
Cmd+/
Toggle the sidebar. Useful when you want max screen real estate.
Cmd+Shift+L
Dark/light mode toggle. Small thing, surprisingly nice.
Shift+Enter
New line without sending. Essential when you're pasting structured prompts with multiple sections.

On Windows or Linux, swap Cmd for Ctrl. ChatGPT and Gemini have similar shortcuts; press ? in each to see their current list.

B · The real shortcuts

Your slash commands

In Claude Code (the terminal tool), slash commands like /clear and /compact are everywhere. For exec-level users, the equivalent is something better: the Skills, Custom GPTs, and Gems you build yourself.

The compounding move
WHY The shift

Generic shortcuts vs. your shortcuts.

Generic shortcuts save 5 seconds per use. Worth knowing, but everyone has them.

Your shortcuts (Skills, Custom GPTs, Gems) save 20-60 minutes per use, get sharper every week, and reflect your specific job, voice, and judgment. Nobody else has them.

This is why Stage 2 is the compounding stage. Every teammate you build is a slash command nobody can copy. Five teammates in your sidebar is a moat.

HOW Where to start

Build your personal shortcut pack.

  1. List the 5 things you do most often. Look at your last 2 weeks. Patterns will show up.
  2. Pick the most painful of the five. Painful = takes too long, output is inconsistent, or you procrastinate it.
  3. Build it as a teammate. See Stage 2 for the recipe.
  4. Use it as your "slash command" for that task. Bookmark it. Pin it. Make it 1 click away.
  5. Repeat for the next 4. By the end of the quarter, you have 5 slash commands no one else has.
C · Reference

The actual / commands

Heads up: these live in Claude Code, Anthropic's terminal tool aimed at developers. They don't work in claude.ai chat. But you asked, so here's the verified list. If you're not in the terminal, skim this and skip to the next section.

Verified May 2026
Session & Context
Manage your conversation
The daily drivers. You'll use these every session.
  • /helpLists every available command in your current session
  • /clearWipe the context and start fresh. Use when switching tasks.
  • /compactSummarize older messages to free up context. Use at ~80% full.
  • /contextSee what's filling your context window right now
  • /costSee your token spend for the session
  • /copy [N]Copy the Nth-latest response to clipboard
  • /exportSave the full conversation as text
Mode & Model
Change how Claude behaves
Switch behavior mid-session without restarting.
  • /planPlan mode: Claude proposes each step, you approve before it acts. Use for unfamiliar codebases or big refactors.
  • /modelSwitch between Opus, Sonnet, Haiku mid-session. Opus for hard reasoning, Sonnet for daily work.
  • /effortSet reasoning depth: low / medium / high / xhigh. Higher = better but slower.
  • /fastToggle Fast Mode for rapid iteration on small fixes
  • /voiceToggle voice input (where supported)
Project & Setup
Configure your environment
Run once per project. Run again when something breaks.
  • /initCreate a CLAUDE.md in this project. Claude reads it every session.
  • /configAdjust settings, permissions, models for the current project
  • /doctorDiagnostic check. First thing to run if anything feels broken.
  • /release-notesSee what shipped recently. Claude Code updates weekly.
  • /keybindingsCustomize your keyboard shortcuts
  • /loginSign in or switch accounts
Extend & Specialize
Add capabilities
Once you're past the basics, these compound.
  • /agentsManage sub-agents. Delegate tasks to specialized helpers.
  • /mcpConnect external tools (GitHub, Slack, databases, etc.) via Model Context Protocol
  • /pluginsInstall and manage community plugins
  • /reviewBundled skill: review code changes for bugs, security, style
  • /debugBundled skill: diagnose what went wrong in this session
  • /loopBundled skill: keep working across turns until a clear completion condition is met
Universal · @ Mentions
Pull files, folders, URLs into your prompt
These also work in claude.ai chat for file references. Use them instead of pasting.
@filename.md Include a single file from your project, full content available to Claude
@src/components/ Include an entire directory, every file inside it
@https://example.com Fetch a URL as read-only context

In any Claude Code session, type / on its own to see your current full list. New commands ship weekly, and plugins or skills add more.

D · Power moves

Workflow habits that compound

Saves 10–20x tokens on long threads. Produces cleaner outputs because the AI is fixing the prompt, not navigating contradictions.

Hit ↑ when the input is empty. Edit. Re-run.

Claude, ChatGPT, and Gemini all natively read PDFs, images, CSVs, and code files. You don't need to convert anything.

Drag a 50-page PDF and ask: "What are the 3 most important things in here for someone in [my role]?"

Take a 10-minute walk. Talk through a decision or a draft. The transcript inserts at the cursor.

Talking it through often surfaces ideas faster than typing. The AI's responses are also better because you're giving it more context than you would in a text prompt.

In Claude: Settings → Styles. Write the rules once. Every chat applies them automatically.

Examples: "Sentence fragments OK. No em-dashes. No exclamation points. Plain English. Hedge upfront." Your voice, made permanent.

Claude and ChatGPT both have Deep Research modes that spend 5-30 minutes synthesizing across many sources. Use it for market scans, competitor research, due diligence prep, technical landscape analysis.

Don't use it for quick answers; it's overkill. Use it when you'd otherwise lose half a day.

When you've had a great thread that led to a good output, share the whole chat. Colleagues can see the reasoning, the back-and-forth, and the final output. Way better than pasting just the answer.

In Claude: chat menu → Share. In ChatGPT: similar. Read-only by default.

E · Power prompts

9 prompts worth saving

Copy these into a notes doc and adapt them. Each card has a copy button.

/ critique

Kill My Idea

Before any board meeting, big email, or strategic decision.

You are a skeptical [board member / customer / competitor]. Find the 3 sharpest objections to this. Push hard. End with the one thing I should test before going further.
/ summarize

Exec Summary

When handed a long doc with 5 min to read.

Summarize this in 5 bullets max. Headline first. What changed. What the 1 thing my [board / team / CFO] will ask. What's missing from this doc that should be in it.
/ rewrite

In My Voice

When output sounds like AI. Pair with 3 past samples.

Rewrite this in my voice based on the samples I've shared. Keep substance identical. Match my sentence length, my use of fragments, my level of hedging. No exclamation points. No "synergy."
/ translate

Audience Shift

Same content, different audience.

Take this doc and rewrite for [audience: board / frontline / 12-year-old / CFO]. Keep substance. Change vocabulary, detail level, assumed knowledge. Same length or shorter.
/ synthesize

Raw → Structured

Customer calls, survey responses, focus groups, town halls.

From this raw input, extract: top 3 themes, strongest quote per theme, what's NOT being said, 5 sharpest questions I should ask next. Flag inferred vs. quoted.
/ coach

Pressure-Test Me

Before a tough conversation or decision. Use with voice mode.

I'm about to [decision / conversation]. Ask the 5 hardest questions a smart skeptic would ask. Don't answer them. After I respond, push back on my answers.
/ structure

Frame the Mess

For complex decisions with 3+ moving parts.

Here's a messy situation: [paste 2–3 paragraphs]. Propose 2 frameworks for thinking about this, with pros/cons of each. Then recommend which fits this specific situation, and why.
/ extract

Action Items

From a meeting transcript or long email thread.

From this transcript/thread, extract: decisions made, action items (with owners and deadlines), open questions, and 3 most important things someone joining late should know.
/ steelman

Best Counterargument

When you're sure you're right.

Steelman the opposite of my position. Make the strongest possible case against my view. Not strawman. Strongest. End with the 1 thing about my view I should most worry about.
Do this

Try this week

  1. Learn 3 keyboard shortcuts. Pick Cmd+K, ↑, and one more. Use them daily for a week. Muscle memory takes ~5 days.
  2. Save 5 power prompts. Copy from above into a notes doc. Title it "My Prompts." Add to it weekly.
  3. Pick your future slash commands. Look at your 5 prompts. Which 1 is painful enough that it deserves to become a Stage 2 teammate? Build it this weekend.
You've reached the end

Back to the start

If something here was useful, send it to a friend. If something was off, tell me. The Shenoy Fund exists to help our community navigate big shifts; this guide is one way we try to do that.

Home →