Five Claude prompt codes
Five short codes you put in front of a prompt to get committed answers, sharper plans, human-sounding writing and deeper reasoning.
How these work
Type the code at the start of your message, then your normal prompt. Claude reads it as an instruction about how to answer.
Worth knowing: four of these are prompt conventions, not built-in commands, so results vary and spelling the intent out always helps. ultrathink is the exception: in Claude Code it is a real keyword that gives Claude its largest thinking budget.
1. L99
Stops the hedging. Claude commits to a real answer instead of "it depends".
L99 Should I launch the paid plan this month or wait until January? Give me one answer and your reasoning.
2. OODA
Runs your situation through the Observe, Orient, Decide, Act loop, a military decision framework: what to look at, what matters, what to decide and the actions to take.
OODA Our biggest client just asked for a 30% discount on renewal. Here's the context: [details]
3. /ghost
Strips the AI-sounding patterns from writing so it reads like a person wrote it.
/ghost Rewrite this LinkedIn post so it sounds like me, not a template: [post]
4. scaffold
Turns a project or task into a full action plan with every step mapped out.
scaffold I want to launch a weekly newsletter for freelance designers by the end of the month.
5. ultrathink
Asks for maximum reasoning before the answer. Use it for hard problems, not quick questions: it is slower and uses more of your limits.
ultrathink Find the flaw in this pricing model and show me the numbers: [model]
Make any code work better
- Pair the code with the outcome you want: "L99, one recommendation, under 100 words."
- Give real context. A code changes how Claude answers, not what it knows about your situation.
- If the result drifts, say what the code means in plain words once. Claude will hold it for the rest of the chat.