These are the ones worth building into your workflow. Use the shorthand if you like, but the plain-English version is what actually does the work.
Shorthand
What it does
Say this instead
L99
Direct, opinion-first answer with no hedging
"Give me your direct recommendation first, then the specific reasoning. Don't hedge or list every option equally."
ULTRATHINK
Deeper reasoning on a hard problem
"Think this through carefully before answering, consider the alternatives, and tell me which you'd choose and why." In the Claude app, also switch on Extended thinking.
/ghost
Rewrites text so it reads like a person wrote it
"Rewrite this in natural, plain language. No clichés, no 'delve', no 'in today's fast-paced world', no em dash habit."
/mirror
Matches your voice from a sample
"Here's a sample of my writing. Match its tone, sentence length and vocabulary in everything you write next."
PERSONA:
Frames Claude as a specific expert
"Answer as a [very specific expert, e.g. Postgres DBA with 15 years running payments infrastructure]." Specific beats generic every time.
/premortem
Finds why your plan will fail before it does
"Assume this plan failed 12 months from now. List the most likely reasons, ranked, and how to prevent each one."
/punch
Makes copy hit harder
"Tighten this copy. Shorter sentences, active voice, strongest point first, cut weak words."
PARETO
Finds the 20% that drives 80% of results
"Which 20% of these actions will drive most of the result? Rank them and tell me what I can drop."
/noyap
Cuts the filler
"Answer directly. No intro, no recap, no closing offer, no filler."
INVERT
Solves the problem backwards
"Work backwards. What would guarantee failure here, and what does avoiding each of those tell us to do?"
✍️ Writing and style
Same deal: these are shorthand, so be specific about what you want changed.
Shorthand
What to ask for
/ghost ⭐
Rewrite so it doesn't read like AI wrote it
/mirror ⭐
Match my writing style from a sample I give you
/punch ⭐
Make every sentence hit harder and more direct
/noyap ⭐
Remove filler, intros and AI-style phrasing
/raw
No formatting, no headings or bullets, just plain text
/voice
Keep this exact tone for the whole conversation
/flow
Restructure so it reads smoothly start to finish
/trim
Cut the fluff without losing meaning
/hook
Rewrite the opening line to grab attention
/rephrase
Say the same thing in a completely different way
/polish
Make rough text sound professional without changing meaning
🧠 Thinking and reasoning
Shorthand
What to ask for
L99 ⭐
Your most specific, opinionated answer
ULTRATHINK ⭐
Think it through properly before answering
/premortem ⭐
Assume the plan already failed and work out why
INVERT ⭐
Solve it backwards from the goal
OODA
Use the observe, orient, decide, act framework
/deepthink
Reason through every layer before answering
CHAINLOGIC
Show each reasoning step so I can follow the logic
/blindspots
Tell me what I'm missing that I didn't think to ask
OVERTHINK
Over-analyse on purpose to catch details others miss
/unpack
Break this idea into every piece that makes it work
XRAY
Look past the obvious answer to find the real one
/layered
Answer at surface, mid and deep levels
/skeptic
Challenge my assumptions and poke holes in the logic
Think step by step
Work through it in stages before giving the answer
📊 Strategy and analysis
Shorthand
What to ask for
PARETO ⭐
The 20% of effort that drives 80% of results
/analyst
Structured, data-backed analysis
/critic
A critical, challenge-everything review
/compare
A side-by-side comparison table
SCAFFOLD / BLUEPRINT
A full project plan with structure and files
/pitch
Frame it as a pitch or persuasive argument
🛠️ Creation and artifacts
In the Claude app, Claude builds interactive things (apps, dashboards, diagrams) as artifacts when you ask for them. You don't need a code. Just say what you want built.
Shorthand
What to ask for
ARTIFACTS
"Build this as an interactive artifact"
/buildme
A complete working app or tool from a description
DASHBOARD
An interactive dashboard from my data
PROTOTYPE
A working prototype of this idea
CANVAS / /render
A visual layout or diagram
WIREFRAME
A basic layout for a website or app screen
GENERATOR
A reusable tool that creates outputs on demand
🎭 Framing and persona
Shorthand
What to ask for
PERSONA: [role] ⭐
Respond as a specific expert. It must be specific to work: "PERSONA: Postgres DBA with 15 years at a payments company", not "PERSONA: developer".
💻 Claude Code: real built-in commands
These are different. They're genuine commands built into Claude Code (the coding tool), and they actually trigger features.
Command
What it does
/clear
Starts a fresh conversation with empty context
/compact
Summarises the conversation so far to free up space
/context
Shows what's filling up the context window
/usage
Shows your session cost and plan usage limits (/cost still works as a shortcut)
/rewind
Rolls the conversation and/or your code back to an earlier point
/resume
Picks up a previous conversation
/model
Switches the model
/effort
Sets how hard Claude thinks for the whole session
/plan
Switches to plan mode so Claude plans before it changes anything
/goal
Sets a finish line. Claude keeps working until it's met
/btw
Asks a side question without adding it to the conversation
/insights
Builds a report on how you use Claude Code and what to try next
/simplify
Reviews your changed code and cleans it up
/code-review
Reviews your changes for bugs
/init
Creates a starter CLAUDE.md file for your project
/memory
Edits your CLAUDE.md memory files
/permissions
Sets what Claude can do without asking
/mcp
Manages your MCP server connections
❌ Don't bother with these
Viral, but not real. Claude has no such modes, so at best they add noise.
BEASTMODE: usually just makes the answer longer
OMEGA: does nothing
/jailbreak: does nothing (if anything it makes Claude more careful)
/kill: does nothing
Follow me for more AI workflows 🤙
Every new guide, tool and setup goes out as a short video first. The write-ups land here.
