10 Secret Claude Codes That Change How You Think and Build

Ugur Aktas

Jul 11, 2026

claudeclaude-codepromptsai-toolsproductivity

10 secret codes for Claude — type one of these words at the start of your prompt and Claude switches into a completely different mode. No settings to change. No custom instructions. Just a keyword.

These aren't gimmicks. Each code triggers a specific reasoning style — simplification, stress-testing, teaching, deep analysis, or production-grade code. Save this list and paste the example prompts when you need them.

Quick Reference

Code	What it does
ELI5	Explains like you're five — strips jargon, uses simple analogies
TLDR	Compresses long text into one paragraph
STEELMAN	Builds the strongest argument against your idea
RED TEAM	Finds weaknesses in your plan from an attacker's perspective
PREMORTEM	Imagines your project already failed — tells you why
SOCRATIC	Teaches by asking questions instead of giving answers
FIVE WHYS	Keeps asking "why" until it hits the root cause
REWRITE AS SENIOR	Rewrites your code to senior developer quality
DEVIL'S ADVOCATE	Deliberately argues against your decision
ULTRATHINK	Enables deep thinking mode in Claude Code for hard problems

1. ELI5 — Explain Like I'm 5

What it does: Takes a complex topic and explains it in plain language, as if talking to a five-year-old. No jargon. Simple analogies. Short sentences.

When to use it: Learning a new concept, explaining tech to a non-technical teammate, or cutting through dense documentation.

Example prompt:

ELI5: How does a database index work and why does it make queries faster?

Another example:

ELI5: What is an API and why do apps need them? Use a real-world analogy.

2. TLDR — Too Long; Didn't Read

What it does: Summarizes a long piece of text into a single tight paragraph. Keeps only what matters.

When to use it: Long articles, meeting notes, legal docs, research papers, or chat threads you don't have time to read fully.

Example prompt:

TLDR: [paste your long text here]

Another example:

TLDR this article for me. I need the key takeaway, who it's for, and one action I should take:

[paste article]

3. STEELMAN — Strongest Counter-Argument

What it does: Instead of attacking your idea with weak objections, Claude builds the strongest possible argument against it — the version a smart opponent would actually use.

When to use it: Before pitching an idea, making a big bet, or publishing something controversial. Forces you to face the best counter-argument, not a straw man.

Example prompt:

STEELMAN: I want to quit my job and go full-time on my SaaS. The product has 200 paying users and $3k MRR. Give me the strongest argument against doing this right now.

Another example:

STEELMAN my decision to migrate from PostgreSQL to MongoDB. Make the counter-argument as convincing as possible.

4. RED TEAM — Attack Your Plan

What it does: Claude puts on an adversary hat and tries to break your plan — security holes, logic gaps, edge cases, things you'd miss because you're too close to the project.

When to use it: Launch plans, security reviews, system architecture, marketing campaigns, anything where blind spots are expensive.

Example prompt:

RED TEAM: We're launching a public API next week. Auth is JWT, rate limit is 100 req/min, data is user profiles. Find every way this could go wrong or get exploited.

Another example:

RED TEAM my onboarding flow. Assume you're a competitor trying to copy our product or a user trying to break it. What weaknesses do you find?

5. PREMORTEM — Assume It Already Failed

What it does: Claude imagines your project has already failed spectacularly, then works backward to explain why. More structured than RED TEAM — focused on project failure, not adversarial attack.

When to use it: Before starting a big project, hiring, fundraising, or any decision with high stakes and irreversible cost.

Example prompt:

PREMORTEM: It's 6 months from now. Our mobile app launch failed. Users downloaded it but retention is under 5%. Tell me exactly what went wrong and what we should have seen coming.

Another example:

PREMORTEM: We hired 3 engineers in 2 months and the team is slower, not faster. Walk me through how we got here.

6. SOCRATIC — Learn by Questions

What it does: Claude stops giving direct answers and starts asking you questions — guiding you to figure it out yourself. Like a tutor, not a search engine.

When to use it: Learning new skills, studying for interviews, understanding why something works, not just what the answer is.

Example prompt:

SOCRATIC: I want to understand recursion in programming. Don't explain it directly — ask me questions that lead me to understand it myself.

Another example:

SOCRATIC: I'm trying to decide between building a monolith or microservices for my startup. Ask me the right questions instead of giving me an answer.

7. FIVE WHYS — Find the Root Cause

What it does: Claude asks "why" repeatedly — usually five times — drilling down from a symptom to the actual root cause. Classic problem-solving technique from Toyota's production system.

When to use it: Debugging business problems, churn analysis, recurring bugs, team conflicts, anything where the obvious answer isn't the real answer.

Example prompt:

FIVE WHYS: Our checkout conversion dropped 30% last month. Start with that symptom and keep asking why until you reach the root cause.

Another example:

FIVE WHYS: Users keep abandoning onboarding at step 3. Dig into the root cause.

8. REWRITE AS SENIOR — Production-Grade Code

What it does: Takes your code and rewrites it the way a senior developer would — better naming, error handling, structure, edge cases, readability. Not just "make it work" — make it maintainable.

When to use it: Code reviews, learning best practices, cleaning up prototypes before shipping, or leveling up your own code.

Example prompt:

REWRITE AS SENIOR:

```typescript
async function getUser(id) {
  const res = await fetch('/api/users/' + id)
  const data = await res.json()
  return data
}
```

Keep the same functionality. Add proper error handling, types, and naming.

Another example:

REWRITE AS SENIOR this React component. Fix performance issues, extract logic, and follow best practices:

[paste your component]

9. DEVIL'S ADVOCATE — Argue the Other Side

What it does: Claude deliberately takes the opposite position on your decision — not to be annoying, but to expose the weak points you haven't considered. Lighter than STEELMAN; more conversational pushback.

When to use it: Decisions you're already leaning toward, team debates, product trade-offs, or when you need someone to challenge your assumptions.

Example prompt:

DEVIL'S ADVOCATE: I think we should add a free tier to increase signups. Convince me this is a bad idea.

Another example:

DEVIL'S ADVOCATE: I want to rebuild our entire frontend in Next.js. Push back on this decision hard.

10. ULTRATHINK — Deep Thinking Mode (Claude Code)

What it does: Activates Claude Code's extended thinking mode. Claude spends more time reasoning before responding — better for architecture decisions, complex bugs, multi-step refactors, and problems where the first answer is usually wrong.

When to use it: Hard technical problems in Claude Code — system design, tricky bugs, performance optimization, security analysis.

Example prompt:

ULTRATHINK: Our Next.js app has intermittent 500 errors under load. The error only happens on Vercel, not locally. Logs show timeout on database queries. Diagnose the root cause and propose a fix.

Another example:

ULTRATHINK: Design the database schema and API architecture for a multi-tenant SaaS with org-level billing, role-based access, and audit logs. Think through edge cases before proposing anything.

Note: ULTRATHINK works in Claude Code (the terminal/IDE agent), not in regular claude.ai chat. Type it at the start of your prompt in Claude Code.

How to Use These

Put the code at the start of your prompt — ELI5:, RED TEAM:, ULTRATHINK:, etc.
Be specific after the code. The trigger sets the mode; your context sets the quality.
Combine when needed. Example: PREMORTEM on a plan, then RED TEAM on the fix.
Save prompts that work. These are starting points — tweak them for your projects.

Start Right Now

Pick the problem you're stuck on. Match it to a code:

Don't understand something → ELI5
Too much to read → TLDR
About to make a big decision → STEELMAN or DEVIL'S ADVOCATE
Launching something → RED TEAM or PREMORTEM
Want to actually learn → SOCRATIC
Something keeps breaking → FIVE WHYS
Code needs to ship → REWRITE AS SENIOR
Hard bug in Claude Code → ULTRATHINK

Copy an example prompt above, swap in your context, and run it.

← Back to blog

10 Secret Claude Codes That Change How You Think and Build

Ugur Aktas

Jul 11, 2026

claudeclaude-codepromptsai-toolsproductivity

Quick Reference

Code	What it does
ELI5	Explains like you're five — strips jargon, uses simple analogies
TLDR	Compresses long text into one paragraph
STEELMAN	Builds the strongest argument against your idea
RED TEAM	Finds weaknesses in your plan from an attacker's perspective
PREMORTEM	Imagines your project already failed — tells you why
SOCRATIC	Teaches by asking questions instead of giving answers
FIVE WHYS	Keeps asking "why" until it hits the root cause
REWRITE AS SENIOR	Rewrites your code to senior developer quality
DEVIL'S ADVOCATE	Deliberately argues against your decision
ULTRATHINK	Enables deep thinking mode in Claude Code for hard problems

1. ELI5 — Explain Like I'm 5

What it does: Takes a complex topic and explains it in plain language, as if talking to a five-year-old. No jargon. Simple analogies. Short sentences.

When to use it: Learning a new concept, explaining tech to a non-technical teammate, or cutting through dense documentation.

Example prompt:

ELI5: How does a database index work and why does it make queries faster?

Another example:

ELI5: What is an API and why do apps need them? Use a real-world analogy.

2. TLDR — Too Long; Didn't Read

What it does: Summarizes a long piece of text into a single tight paragraph. Keeps only what matters.

When to use it: Long articles, meeting notes, legal docs, research papers, or chat threads you don't have time to read fully.

Example prompt:

TLDR: [paste your long text here]

Another example:

TLDR this article for me. I need the key takeaway, who it's for, and one action I should take:

[paste article]

3. STEELMAN — Strongest Counter-Argument

What it does: Instead of attacking your idea with weak objections, Claude builds the strongest possible argument against it — the version a smart opponent would actually use.

When to use it: Before pitching an idea, making a big bet, or publishing something controversial. Forces you to face the best counter-argument, not a straw man.

Example prompt:

STEELMAN: I want to quit my job and go full-time on my SaaS. The product has 200 paying users and $3k MRR. Give me the strongest argument against doing this right now.

Another example:

STEELMAN my decision to migrate from PostgreSQL to MongoDB. Make the counter-argument as convincing as possible.

4. RED TEAM — Attack Your Plan

What it does: Claude puts on an adversary hat and tries to break your plan — security holes, logic gaps, edge cases, things you'd miss because you're too close to the project.

When to use it: Launch plans, security reviews, system architecture, marketing campaigns, anything where blind spots are expensive.

Example prompt:

RED TEAM: We're launching a public API next week. Auth is JWT, rate limit is 100 req/min, data is user profiles. Find every way this could go wrong or get exploited.

Another example:

RED TEAM my onboarding flow. Assume you're a competitor trying to copy our product or a user trying to break it. What weaknesses do you find?

5. PREMORTEM — Assume It Already Failed

When to use it: Before starting a big project, hiring, fundraising, or any decision with high stakes and irreversible cost.

Example prompt:

PREMORTEM: It's 6 months from now. Our mobile app launch failed. Users downloaded it but retention is under 5%. Tell me exactly what went wrong and what we should have seen coming.

Another example:

PREMORTEM: We hired 3 engineers in 2 months and the team is slower, not faster. Walk me through how we got here.

6. SOCRATIC — Learn by Questions

What it does: Claude stops giving direct answers and starts asking you questions — guiding you to figure it out yourself. Like a tutor, not a search engine.

When to use it: Learning new skills, studying for interviews, understanding why something works, not just what the answer is.

Example prompt:

SOCRATIC: I want to understand recursion in programming. Don't explain it directly — ask me questions that lead me to understand it myself.

Another example:

SOCRATIC: I'm trying to decide between building a monolith or microservices for my startup. Ask me the right questions instead of giving me an answer.

7. FIVE WHYS — Find the Root Cause

What it does: Claude asks "why" repeatedly — usually five times — drilling down from a symptom to the actual root cause. Classic problem-solving technique from Toyota's production system.

When to use it: Debugging business problems, churn analysis, recurring bugs, team conflicts, anything where the obvious answer isn't the real answer.

Example prompt:

FIVE WHYS: Our checkout conversion dropped 30% last month. Start with that symptom and keep asking why until you reach the root cause.

Another example:

FIVE WHYS: Users keep abandoning onboarding at step 3. Dig into the root cause.

8. REWRITE AS SENIOR — Production-Grade Code

When to use it: Code reviews, learning best practices, cleaning up prototypes before shipping, or leveling up your own code.

Example prompt:

REWRITE AS SENIOR:

```typescript
async function getUser(id) {
  const res = await fetch('/api/users/' + id)
  const data = await res.json()
  return data
}
```

Keep the same functionality. Add proper error handling, types, and naming.

Another example:

REWRITE AS SENIOR this React component. Fix performance issues, extract logic, and follow best practices:

[paste your component]

9. DEVIL'S ADVOCATE — Argue the Other Side

When to use it: Decisions you're already leaning toward, team debates, product trade-offs, or when you need someone to challenge your assumptions.

Example prompt:

DEVIL'S ADVOCATE: I think we should add a free tier to increase signups. Convince me this is a bad idea.

Another example:

DEVIL'S ADVOCATE: I want to rebuild our entire frontend in Next.js. Push back on this decision hard.

10. ULTRATHINK — Deep Thinking Mode (Claude Code)

When to use it: Hard technical problems in Claude Code — system design, tricky bugs, performance optimization, security analysis.

Example prompt:

ULTRATHINK: Our Next.js app has intermittent 500 errors under load. The error only happens on Vercel, not locally. Logs show timeout on database queries. Diagnose the root cause and propose a fix.

Another example:

ULTRATHINK: Design the database schema and API architecture for a multi-tenant SaaS with org-level billing, role-based access, and audit logs. Think through edge cases before proposing anything.

Note: ULTRATHINK works in Claude Code (the terminal/IDE agent), not in regular claude.ai chat. Type it at the start of your prompt in Claude Code.

How to Use These

Put the code at the start of your prompt — ELI5:, RED TEAM:, ULTRATHINK:, etc.
Be specific after the code. The trigger sets the mode; your context sets the quality.
Combine when needed. Example: PREMORTEM on a plan, then RED TEAM on the fix.
Save prompts that work. These are starting points — tweak them for your projects.

Start Right Now

Pick the problem you're stuck on. Match it to a code:

Don't understand something → ELI5
Too much to read → TLDR
About to make a big decision → STEELMAN or DEVIL'S ADVOCATE
Launching something → RED TEAM or PREMORTEM
Want to actually learn → SOCRATIC
Something keeps breaking → FIVE WHYS
Code needs to ship → REWRITE AS SENIOR
Hard bug in Claude Code → ULTRATHINK

Copy an example prompt above, swap in your context, and run it.

Quick Reference

Code	What it does
ELI5	Explains like you're five — strips jargon, uses simple analogies
TLDR	Compresses long text into one paragraph
STEELMAN	Builds the strongest argument against your idea
RED TEAM	Finds weaknesses in your plan from an attacker's perspective
PREMORTEM	Imagines your project already failed — tells you why
SOCRATIC	Teaches by asking questions instead of giving answers
FIVE WHYS	Keeps asking "why" until it hits the root cause
REWRITE AS SENIOR	Rewrites your code to senior developer quality
DEVIL'S ADVOCATE	Deliberately argues against your decision
ULTRATHINK	Enables deep thinking mode in Claude Code for hard problems

1. ELI5 — Explain Like I'm 5

What it does: Takes a complex topic and explains it in plain language, as if talking to a five-year-old. No jargon. Simple analogies. Short sentences.

When to use it: Learning a new concept, explaining tech to a non-technical teammate, or cutting through dense documentation.

Example prompt:

ELI5: How does a database index work and why does it make queries faster?

Another example:

ELI5: What is an API and why do apps need them? Use a real-world analogy.

2. TLDR — Too Long; Didn't Read

What it does: Summarizes a long piece of text into a single tight paragraph. Keeps only what matters.

When to use it: Long articles, meeting notes, legal docs, research papers, or chat threads you don't have time to read fully.

Example prompt:

TLDR: [paste your long text here]

Another example:

TLDR this article for me. I need the key takeaway, who it's for, and one action I should take:

[paste article]

3. STEELMAN — Strongest Counter-Argument

What it does: Instead of attacking your idea with weak objections, Claude builds the strongest possible argument against it — the version a smart opponent would actually use.

When to use it: Before pitching an idea, making a big bet, or publishing something controversial. Forces you to face the best counter-argument, not a straw man.

Example prompt:

STEELMAN: I want to quit my job and go full-time on my SaaS. The product has 200 paying users and $3k MRR. Give me the strongest argument against doing this right now.

Another example:

STEELMAN my decision to migrate from PostgreSQL to MongoDB. Make the counter-argument as convincing as possible.

4. RED TEAM — Attack Your Plan

What it does: Claude puts on an adversary hat and tries to break your plan — security holes, logic gaps, edge cases, things you'd miss because you're too close to the project.

When to use it: Launch plans, security reviews, system architecture, marketing campaigns, anything where blind spots are expensive.

Example prompt:

RED TEAM: We're launching a public API next week. Auth is JWT, rate limit is 100 req/min, data is user profiles. Find every way this could go wrong or get exploited.

Another example:

RED TEAM my onboarding flow. Assume you're a competitor trying to copy our product or a user trying to break it. What weaknesses do you find?

5. PREMORTEM — Assume It Already Failed

When to use it: Before starting a big project, hiring, fundraising, or any decision with high stakes and irreversible cost.

Example prompt:

PREMORTEM: It's 6 months from now. Our mobile app launch failed. Users downloaded it but retention is under 5%. Tell me exactly what went wrong and what we should have seen coming.

Another example:

PREMORTEM: We hired 3 engineers in 2 months and the team is slower, not faster. Walk me through how we got here.

6. SOCRATIC — Learn by Questions

What it does: Claude stops giving direct answers and starts asking you questions — guiding you to figure it out yourself. Like a tutor, not a search engine.

When to use it: Learning new skills, studying for interviews, understanding why something works, not just what the answer is.

Example prompt:

SOCRATIC: I want to understand recursion in programming. Don't explain it directly — ask me questions that lead me to understand it myself.

Another example:

SOCRATIC: I'm trying to decide between building a monolith or microservices for my startup. Ask me the right questions instead of giving me an answer.

7. FIVE WHYS — Find the Root Cause

What it does: Claude asks "why" repeatedly — usually five times — drilling down from a symptom to the actual root cause. Classic problem-solving technique from Toyota's production system.

When to use it: Debugging business problems, churn analysis, recurring bugs, team conflicts, anything where the obvious answer isn't the real answer.

Example prompt:

FIVE WHYS: Our checkout conversion dropped 30% last month. Start with that symptom and keep asking why until you reach the root cause.

Another example:

FIVE WHYS: Users keep abandoning onboarding at step 3. Dig into the root cause.

8. REWRITE AS SENIOR — Production-Grade Code

When to use it: Code reviews, learning best practices, cleaning up prototypes before shipping, or leveling up your own code.

Example prompt:

REWRITE AS SENIOR:

```typescript
async function getUser(id) {
  const res = await fetch('/api/users/' + id)
  const data = await res.json()
  return data
}
```

Keep the same functionality. Add proper error handling, types, and naming.

Another example:

REWRITE AS SENIOR this React component. Fix performance issues, extract logic, and follow best practices:

[paste your component]

9. DEVIL'S ADVOCATE — Argue the Other Side

When to use it: Decisions you're already leaning toward, team debates, product trade-offs, or when you need someone to challenge your assumptions.

Example prompt:

DEVIL'S ADVOCATE: I think we should add a free tier to increase signups. Convince me this is a bad idea.

Another example:

DEVIL'S ADVOCATE: I want to rebuild our entire frontend in Next.js. Push back on this decision hard.

10. ULTRATHINK — Deep Thinking Mode (Claude Code)

When to use it: Hard technical problems in Claude Code — system design, tricky bugs, performance optimization, security analysis.

Example prompt:

ULTRATHINK: Our Next.js app has intermittent 500 errors under load. The error only happens on Vercel, not locally. Logs show timeout on database queries. Diagnose the root cause and propose a fix.

Another example:

ULTRATHINK: Design the database schema and API architecture for a multi-tenant SaaS with org-level billing, role-based access, and audit logs. Think through edge cases before proposing anything.

Note: ULTRATHINK works in Claude Code (the terminal/IDE agent), not in regular claude.ai chat. Type it at the start of your prompt in Claude Code.

How to Use These

Put the code at the start of your prompt — ELI5:, RED TEAM:, ULTRATHINK:, etc.
Be specific after the code. The trigger sets the mode; your context sets the quality.
Combine when needed. Example: PREMORTEM on a plan, then RED TEAM on the fix.
Save prompts that work. These are starting points — tweak them for your projects.

Start Right Now

Pick the problem you're stuck on. Match it to a code:

Don't understand something → ELI5
Too much to read → TLDR
About to make a big decision → STEELMAN or DEVIL'S ADVOCATE
Launching something → RED TEAM or PREMORTEM
Want to actually learn → SOCRATIC
Something keeps breaking → FIVE WHYS
Code needs to ship → REWRITE AS SENIOR
Hard bug in Claude Code → ULTRATHINK

Copy an example prompt above, swap in your context, and run it.