The 5 Claudeprompt codes

The 5 codes from the video, and how to install them as skills so Claude runs them properly every time. No coding, no terminal.

The 5 codes

Type any of these into Claude with the thing you want it to look at. They work right now, on any plan, before you install anything.

1. Run a pre-mortem on this.
2. Red team this.
3. Steel man the other side.
4. Explain this like I'm 5.
5. Break this down from first principles.

That is the short version and it works. The rest of this page is the better version.

Why turn them into skills

A code typed into the chat is a suggestion. Claude does its best with one line and then forgets you ever said it.

A skill is the full set of instructions behind that line, saved to your account. Instead of 6 words, Claude gets a page telling it exactly how to run a pre-mortem: start from failure, rank the causes, name the early warning sign, do not soften the conclusion. Same idea, far better answer, and you never type it again.

Think of it as the difference between telling someone “review this” and handing them a checklist. A skill is just a text file. You do not need to code and you will not touch a terminal.

Before you start

You need a paid Claude plan to upload skills: Pro, Max, Team or Enterprise. On the free plan you can still type the 5 codes, you just cannot save them.

You need one setting switched on, which is step 1 below. Nearly everyone who gets stuck is stuck on that setting.

Step 1: Switch on the setting

Go to claude.ai, click your name at the bottom left, then Settings, then Capabilities. Turn on code execution and file creation.

Do this first. Until it is on, the Skills menu does not appear anywhere in the app.

You’ll know it worked when a Skills option appears in that same Capabilities section.

Step 2: Get the skill file

The easy way: let Claude build it

Open a new Claude chat and paste this:

Build me a Claude skill and give me the finished zip to download.

Create a folder called pre-mortem with a single file inside it called SKILL.md. Put the exact text I have pasted below into that file, unchanged. Then zip the folder itself (not just the file) and give me the zip as a download link.

Here is the text for SKILL.md:

[paste the pre-mortem block from the bottom of this page here]

Claude builds it and hands you a zip. Download it. Repeat for the other 4, or ask it to do all 5 at once and give you 5 separate zips.

The manual way

1. On Mac, open TextEdit, then Format menu, Make Plain Text. On Windows, open Notepad.
2. Paste in one of the 5 blocks from the bottom of this page.
3. Save it as SKILL.md. On Mac, untick “if no extension is provided, use .txt”.
4. Make a new folder named the same as the skill, for example pre-mortem, and put the file in it.
5. Right click the folder. On Mac choose Compress. On Windows choose Send to, then Compressed (zipped) folder.

Zip the folder, not the file on its own. This is the single most common mistake and the upload just fails.

Step 3: Upload it

1. Settings, then Capabilities, then Skills. Some accounts show it under Customize, then Skills.
2. Click the plus, then Upload skill.
3. Pick your zip.
4. Claude reads it and shows you a summary of what the skill does.
5. Toggle it on.

You’ll know it worked when the skill sits in your list with a description, switched on.

Step 4: Use it

You do not type a slash. Once a skill is on, Claude reads the descriptions and fires the right one when the job matches. So you just talk normally:

Here is my plan for [thing]. Pre-mortem it.

If you want to be certain it fired, name it:

Use the pre-mortem skill on this.

One rule that decides everything: give it something real. Paste the actual plan, the actual contract, the actual pricing page. These codes sharpen thinking about a thing. With no thing, you get generic output back.

The 5 skill files

Copy each block exactly, including the 3 dashes at the top and bottom of the first section. That top part is what tells Claude when to use the skill.

1. Pre-mortem

---
name: pre-mortem
description: Use whenever the user asks to pre-mortem a plan, idea, launch or decision, asks why something will fail, or says "pretend this failed".
---

# Pre-mortem

Assume the plan has already failed 6 months from now. Work backwards from there.

1. List the failure paths, ranked by how likely each one is. Be specific to what the user wrote, never generic business advice.
2. For each one, name the earliest warning sign they would have seen.
3. Name the single change that removes the most risk.

Do not grade the plan. Do not soften the conclusion. Do not end on reassurance.

2. Red team

---
name: red-team
description: Use whenever the user asks to red team something, poke holes in it, or find where people will push back on a plan, pitch, price or piece of copy.
---

# Red team

Attack the work from 3 angles: a sceptical investor, a direct competitor, and the customer being sold to.

1. Give the 3 hardest questions each of them would ask, in the words they would use.
2. For each question, say whether the user currently has an answer.
3. End with the single question they cannot answer yet.

Do not be polite about it. Do not balance the criticism with praise.

3. Steel man

---
name: steel-man
description: Use whenever the user asks to steel man something, argue the other side, or hear the strongest case against their position or decision.
---

# Steel man

Argue the strongest possible version of the case against the user's position, the one a brilliant person who disagrees would make.

1. State their position back in one line, fairly.
2. Make the opposing case properly, using their own facts where possible.
3. Name the one piece of evidence that would settle it.

Never strawman. Never end by telling them they were right after all.

4. Explain like I’m 5

---
name: eli5
description: Use whenever the user asks to explain something simply, says ELI5, says they do not understand something, or pastes a contract, policy, lease, tax letter or technical document and wants it in plain words.
---

# Explain like I'm 5

Explain it in plain words. No jargon, no analogies that need explaining themselves, short sentences.

Then, separately:
1. The 3 things in here that could cost money.
2. Anything the user would regret agreeing to.
3. Anything written vaguely enough to be read 2 ways.

Keep every number exactly as written in the source. Never estimate.

5. First principles

---
name: first-principles
description: Use whenever the user asks to break something down from first principles, asks to ignore the conventional advice, or says the standard answer is not good enough.
---

# First principles

Strip the question back to what is actually, verifiably true, then rebuild the answer from there.

1. List the assumptions in the question, separating what is genuinely true from what is only industry convention.
2. Drop the inherited ones and say why.
3. Rebuild the answer from the facts that survived.

Flag clearly which assumptions were dropped. Do not restate common advice.

Stacking them

The order that works on anything that matters:

1. First principles, to work out what you are actually solving.
2. Build the plan.
3. Pre-mortem, to find the failure paths.
4. Red team, to find the questions you will be asked.
5. Fix it, then steel man your final call before you commit.

Do it in one conversation so each step sees the last. Do not run all 5 on a small decision. That is a lot of thinking for choosing a font.

If you use Claude Code

Skip the zip entirely. Drop each folder into ~/.claude/skills/ and it is live next session. Same files, no upload.

Troubleshooting

There is no Skills menu. Code execution and file creation is switched off, or you are on the free plan, which cannot upload skills.

The upload was rejected. You zipped the file instead of the folder. Put SKILL.md inside a folder first, then zip the folder.

The file saved as SKILL.md.txt. TextEdit added the extension. Save again with the .txt option unticked, or rename the file.

The skill uploaded but never fires. The description is too vague. Copy the description lines above exactly.

The answer is still generic. You did not paste the real thing. These work on specifics, not a description of a situation.

It agreed with me anyway. Check the skill actually fired by naming it directly. If it did, you are giving it too little to work with.

Nothing happens when I type a slash. On claude.ai skills fire on their own. There is no slash to type. Just describe the job.