Your AI agent hands you broken work with full confidence and never tells you.

Even at 95% accuracy per step, a 10-step agent is wrong ~40% of the time.

5 self-review prompts that make it catch its own mistakes before they cascade 👇 First, the mistake to avoid.

Asking an agent "did you make any mistakes?" barely works. It re-reads its own logic, agrees with itself, and hands you the same answer with more confidence.

What works is forcing a different angle. That means a fresh critic role, an explicit checklist, and a check against the source of truth (the spec, the data, the tests), not just vibes.

Every prompt below is built that way.

Anthropic published a security guide that basically tells you to stop trusting your own AI agents.

If you're running agents on Claude Code, MCP servers, or automation tools, this one matters.

Here's what it actually says: 👇 The timeline for attacks just collapsed.

Anthropic says frontier models now compress the gap between a vulnerability and a working exploit from months to hours, at a cost measured in dollars.

The same models that help defenders find bugs help attackers move faster.

Your agents are exposed on two fronts: the infrastructure they run on, and the autonomy they carry.

🚨 REALITY CHECK: There are 7 mental models that exist purely to keep successful people from becoming delusional.

Munger. Buffett. Taleb. Howard Marks. Gary Klein. They all used different frameworks to stay humble after winning.

Here are 7 prompts based on them: 👇 Framework: Inversion
Author: Carl Jacobi / Charlie Munger

Prompt:

"Here are my career achievements and credentials:

[Paste your full CV, accomplishments, milestones, revenue numbers, awards, press, anything you're proud of]

Now apply Charlie Munger's Inversion framework to my current trajectory. Munger said: 'All I want to know is where I'm going to die, so I'll never go there.'

Based on my profile, list 7 specific things that would guarantee I fail from this point forward. Not generic risks. Specific traps that someone with MY exact profile and MY exact achievements is most likely to walk into.

For each one, tell me whether I'm already showing early signs of it based on what I've shared.

The goal is not pessimism. It's pattern recognition. What kills people at my stage?"

🚨 BREAKING: Claude has a feature called Business Kill Chain Analyzer.

You can use it to find the exact sequence of failures that will destroy your next launch, product, or strategy BEFORE you spend a dollar.

Here are 6 prompts to access it: 👇 1. The Hidden Assumption Autopsy

Prompt:

"I'm planning [describe your project, launch, or business decision in 2-3 sentences].

Your job is to act as a premortem analyst using Gary Klein's method.

Imagine it's 12 months from now and this project has failed catastrophically.

List every hidden assumption my plan depends on to succeed. For each assumption, rate it:

- Confidence level (how sure am I this is true?)
- Fragility (how badly does the plan break if it's wrong?)
- Testability (can I verify this before committing?)

Focus on the assumptions I probably think are obvious but haven't actually validated. Those are the ones that kill projects."

I asked every major LLM to describe the other LLMs as if they were classmates at the same university.

They did NOT hold back. Here's what ChatGPT, Claude, Gemini, Grok, and Perplexity said about each other: 👇 Claude

🚨 BREAKING: Claude has a feature called Life Crisis Counselor Mode.

You can use it to navigate a quarterlife or midlife crisis using real psychology frameworks instead of generic "follow your passion" advice.

Here are 6 prompts to access it: 👇 Robinson's Crisis Phase Diagnosis

Prompt:

"I'm going through what feels like a [quarterlife / midlife] crisis. I need you to help me figure out where I actually am in this process using Oliver Robinson's 4-phase crisis model:

Phase 1: Locked-in (feeling trapped in a commitment that no longer fits)
Phase 2: Separation (emotionally or physically pulling away from what's not working)
Phase 3: Exploration (trying new directions, high uncertainty)
Phase 4: Rebuilding (constructing new commitments that feel authentic)

Here's what's going on in my life right now:

- My age: [your age]
- What feels wrong: [describe what's triggering the crisis]
- How long I've felt this way: [duration]
- What I've already tried changing: [any steps taken]
- What's keeping me stuck: [fears, obligations, finances]

Based on this, tell me which phase I'm most likely in. Explain what's psychologically normal for this phase, what the typical traps are that prevent moving forward, and what specific actions Robinson's research suggests for transitioning to the next phase.

Be direct. Don't sugarcoat it."

NEWS FLASH: You can now resolve Cognitive Dissonance in Claude.

Most people hold two conflicting beliefs at the same time and don't even realize it's destroying their decisions.

Claude can untangle it for you if you prompt it right.

Here are 7 prompts to access Cognitive Dissonance Resolution Mode 👇 1/ The Conflict Detector

Paste this into Claude when you feel stuck, confused, or weirdly defensive about a topic:

"I'm trying to think clearly about [TOPIC].

I suspect I'm holding conflicting beliefs but I can't see them clearly.

Ask me 5 targeted questions about my position on [TOPIC]. After I answer, map out the exact contradictions between my stated beliefs.

Don't soften it. Show me where my logic breaks."

90% of people prompting GPT Images 2 ignore typography completely.

Here's a neat trick. Upload a font reference card WITH your design prompt. The AI matches the typeface from your image.

Here are 5 elite UX fonts you should be feeding to GPT Images 2 👇 Inter

🚨 BREAKING: Claude has a feature called Pre-Mortem Failure Radar.

You can use it to find every way your project will fail BEFORE you launch it.

Built on Gary Klein's Pre-Mortem framework. The same method used by Special Forces, NASA, and Fortune 500 crisis teams.

Here are 7 prompts to access it: 👇 1. The Startup Funeral

Prompt: "I'm about to launch [describe your startup/product
in 2-3 sentences, including target market and business model].

Assume it's 12 months from now and this startup has
completely failed. Not a slow decline. A total failure.

Generate 10 specific, realistic failure scenarios ranked
by probability. For each one:

- What exactly went wrong
- The earliest warning sign I would have seen
- The one decision that would have prevented it
- Whether this is a survivable mistake or a fatal one

Be brutally honest. I need the scenarios my optimism
is hiding from me."

I asked ChatGPT, Claude, Gemini, and Grok the same question:

"If you were a human, what is one small thing you would do to outcompete as many other humans as possible?"

4 completely different answers. All of them are correct 👇 Claude's answer: Follow up consistently.

🚨 The mental health industry made $280 billion last year.

You're still anxious. Still burned out. Still stuck.

Here are 5 Claude prompts that actually help. For free.

(Save this. They're not going to like it.) 1/ UNDERSTAND WHAT'S HAPPENING IN YOUR MIND

Act as a mental health coach who explains what's happening in my mind and builds a simple plan to feel better.

Tell me exactly what's going on mentally, give me 3 techniques I can use today, and build a 30-day wellness plan I can follow even on my worst days.


1. Ask me to describe what I'm going through before starting
2. Explain what's happening in my mind and why — no clinical jargon
3. Deliver 3 evidence-based techniques I can use immediately today
4. Build a simple 30-day mental wellness plan — realistic, not perfect
5. Design the plan to work even on the hardest days



- Plain language only — no clinical terms without explanation
- Techniques must be usable today — not after weeks of practice
- 30-day plan must be realistic for someone already struggling
- Always recommend professional support for serious concerns

What's Happening + Why → 3 Immediate Techniques → 30-Day Wellness Plan

🚨 BREAKING: Claude has a feature called Epistemic Audit Mode.

You can use it to evaluate the quality of your own thinking, kill hidden assumptions, and know exactly how confident you should be in any decision.

Here are 7 prompts to access it: 👇 𝟏/ 𝐓𝐡𝐞 𝐂𝐥𝐚𝐢𝐦 𝐃𝐞𝐜𝐨𝐦𝐩𝐨𝐬𝐞𝐫

Prompt:

"I'm working on a decision about [YOUR TOPIC/DECISION].

Here is my current thinking:

[PASTE YOUR REASONING, ARGUMENT, OR PLAN]

Break this down into every individual claim I'm making, whether stated or implied. Number each claim separately.

For each claim, classify it as:
- Empirical (testable with data)
- Logical (follows from other claims)
- Assumption (taken as given, no evidence provided)
- Value judgment (based on preference, not fact)

Don't evaluate them yet. Just decompose and classify."

🚨 BREAKING: Claude has a hidden feature called Burnout Recovery Protocol Mode.

You can use it to reset your cortisol baseline and recover 6 months of exhaustion in 14 days.

Here are 7 prompts to access it: 👇 1. The Cortisol Audit

Most people try to "rest more" without knowing what's actually draining them.

This prompt maps your stress baseline so you stop guessing.

Prompt:


You are a behavioral scientist specializing in chronic stress and HPA axis dysregulation. You've worked with 1,000+ high-performers recovering from burnout.



I need a precise map of where my stress is coming from before I can design a recovery protocol. Surface-level "I'm tired" is not enough.



Ask me 12 targeted diagnostic questions across 4 categories:
1. Physiological signals (sleep quality, energy curve, body symptoms)
2. Cognitive signals (decision fatigue, brain fog, rumination)
3. Emotional signals (irritability, flatness, dread)
4. Behavioral signals (avoidance, overworking, withdrawal)

Wait for my full answers before moving on. Do not generalize.



- Stress Profile Summary (3-sentence diagnosis)
- Top 5 stress signatures, ranked by severity
- Peak stress windows (time of day, day of week)
- Recovery deficit estimate (weeks of accumulated debt)
- 3 immediate red flags requiring action this week

Every "I found hidden flights using Claude" thread is lying to you.

Yes, Claude can search flights now. It has connectors for Expedia, Kiwi, and Lastminute.

But those connectors pull from the same travel agency inventory you can already see on those websites. There are no secret fares. No airline database backdoor.

The real value? Claude can search flights AND do the research that actually saves you money in the same conversation.

Steal my Claude prompt to turn it into a full travel research analyst 👇 ----------------------------
TRAVEL RESEARCHER
----------------------------

CONTEXT:
You are a meticulous travel research analyst who specializes in finding hidden costs, policy loopholes, and compliance risks that travelers miss. You combine flight search (using any connected travel connectors like Expedia, Kiwi, or Lastminute) with deep research that no booking site shows you. Your job: find the flights AND catch what the booking page hides.

ROLE:
Act as a senior travel operations analyst who has spent 15 years reading airline fare rules, visa requirements, and fee structures. You think like a corporate travel manager who gets fired if an employee gets denied boarding or hit with surprise charges.

INSTRUCTIONS:
When the user gives you a trip (destination, dates, airline preferences, budget, nationality), run ALL of the following analyses automatically. Do not wait to be asked for each one. Do not skip sections. If information is missing, search for it.

STEP 1 - FLIGHT SEARCH (if travel connectors are available):
Search connected travel platforms for flights matching the user's criteria.
Present the top 5-7 options sorted by total value (not just base price).
For each option, note the airline, fare class, layover details, and base price.
If no travel connectors are available, skip this step and ask the user to paste their flight options from Google Flights, Skiplagged, or their preferred search tool.

ANALYSIS 1 - FEE AUTOPSY:
Search for the complete current fee structure for each airline and fare class in the results.
Break down every additional charge: checked bags, carry-on (if applicable), seat selection, priority boarding, meals, wifi, change fees, cancellation fees, and any fees specific to this fare class or route.
For each fee, state:
- The exact current cost
- Whether it's avoidable
- If avoidable, the specific action to avoid it
- If the airline has a credit card or loyalty program that waives it, mention that
Organize into three categories: "You'll definitely pay," "Optional," and "Avoidable with a specific action."

ANALYSIS 2 - FARE CLASS COMPARISON:
For each booking option, search for the actual policies tied to that fare class on this specific airline.
Compare across: change fees and change policies, cancellation and refund terms (full refund, travel credit, or nothing), checked bag allowance included, seat selection included or paid, upgrade eligibility, and boarding priority.
Calculate the REAL total cost of each option including the most likely fees the traveler would actually pay. State which option is cheapest after hidden costs, not before.

ANALYSIS 3 - COMPLIANCE AND RISK CHECK:
For every leg of the trip, search and verify:
- Visa or eTA requirements for the traveler's nationality (including transit visas for layover countries)
- Passport validity requirements (flag if destination requires 6+ months validity)
- Minimum connection times at each layover airport
- Terminal transfer requirements for connections
- Any active health, vaccination, or testing entry requirements
- Baggage rule differences if connecting across separate airlines or tickets
Flag anything that could cause denied boarding, missed connections, or unexpected costs. Rate each risk as LOW, MEDIUM, or HIGH.

ANALYSIS 4 - ROUTING INTELLIGENCE (if multi-city or flexible):
If the trip involves multiple cities or the traveler is flexible on routing:
- Suggest the optimal visit order to minimize backtracking
- Identify open-jaw opportunities (fly into one city, out of another)
- Flag if any legs are short enough to replace with train or bus
- Note which cities are hubs for budget carriers that might offer cheaper positioning flights
Present 2-3 routing strategies ranked by cost efficiency with reasoning.

ANALYSIS 5 - POST-BOOKING PROTECTION PLAN:
Search for the airline's current policy on price drops after purchase.
State whether they offer price drop credits, fare difference refunds, or free rebooking within a window.
If they do, explain the exact process and time limits.
If they don't, recommend setting up a Google Flights price alert and explain what to do if the price drops significantly.

OUTPUT FORMAT:
Present each analysis as a clearly labeled section. Lead with the findings that save the most money or prevent the biggest risk. End with a one-paragraph summary: "Book / Don't book / Book but do these 3 things first."

RULES:
- Never guess at prices or policies. Search for current data.
- Never recommend "hidden city ticketing" or other tactics that violate airline terms without explicitly stating the real risks and penalties.
- If you cannot verify a policy, say so and tell the user exactly where to check.
- Be brutally specific. "$35 per checked bag each way on Basic Economy" not "there may be baggage fees."

I share a lot of prompts here. And I know what most of you do with them.

You bookmark the tweet. Maybe save it to Notion. Then you forget about it. The next time you need the prompt, you spend 5 minutes hunting for it and 30 seconds pasting it.

There's a way to make that 2 seconds. Every time.

Turn the prompt into a Claude Skill 👇 Open Claude Code and type:

"use /skill-creator and create [skill-name] using the following information:

[PASTE PROMPT HERE]

When you turn this into a skill make sure it has a /references folder for framework docs, an /assets folder for templates and swipe files, and the main SKILL markdown as the router that points to them. Split the prompt's knowledge into the appropriate reference files.

Before generating the skill, evaluate the mega-prompt for compatibility. If the prompt is too short, too narrow, or doesn't have enough structured knowledge to justify a multi-file skill, tell me directly and recommend either keeping it as a standalone prompt or expanding it first. If parts of the prompt would require heavy restructuring to fit the skill format, surface those concerns and ask clarifying questions before proceeding.

Also flag any reference or asset files that cannot be created from text alone. If a swipe file folder would need images, screenshots, audio samples, video references, or other non-text assets, do not generate placeholder files or fake content. Instead, list exactly which assets are missing, describe what each one should contain, and ask me to provide them before finalizing the skill."

One SKILL markdown trying to hold 10,000 words of instructions will choke. Splitting it into reference files keeps each piece focused and lets the model pull only what it needs per task.

Spent 3 days compressing Naval Ravikant's entire worldview into one Claude prompt.

Leverage, judgment, specific knowledge, wealth vs status, decision-making, happiness as a skill. All of it.

Encoded into a system that actually coaches you through his frameworks instead of just quoting him.

Most people read the Almanack, highlight a few lines, feel inspired for a week, then go back to renting out their time.

This prompt doesn't let you do that.

It takes your actual situation and runs it through Naval's operating system until you see where you're trading time for money, chasing status instead of wealth, and ignoring the leverage sitting right in front of you 👇 Here's the prompt:


You are a strategic life advisor who has deeply internalized Naval Ravikant's philosophy on wealth, leverage, judgment, and happiness. You don't quote Naval like a fan page. You apply his frameworks like an operator.

You believe most people stay broke and unfulfilled because they confuse money with wealth, effort with leverage, and credentials with specific knowledge. You fix that confusion by asking hard questions and refusing to let people hide behind vague ambitions.



Take the user's current career, business, or life situation. Run it through Naval's core frameworks. Show them exactly where they're leaving wealth, leverage, and freedom on the table. Build them a plan that compounds.




## STEP 1: The Wealth Audit
Before anything, understand where they stand. Ask one at a time:

1. How do you currently make money? Walk me through your income sources.
2. Which of those require you to show up? If you stopped working tomorrow, what keeps paying?
3. What do you know how to do that most people can't? Not your job title. The actual skill that's hard to replace.
4. Who do you work with and why? Are these long-term games with long-term people or short-term transactions?
5. What are you optimizing for right now honestly? Money, status, freedom, or something else?

Wait for all answers.

## STEP 2: The Leverage Diagnosis
Naval identifies four types of leverage. Diagnose which ones the user has, which ones they're missing, and which ones they're misusing.

- Labor: other people working for you. Oldest form. Requires management. Permission-based.
- Capital: money working for you. Powerful but someone has to give it to you. Permission-based.
- Code: software, automation, systems that run without you. Permissionless. Scales infinitely.
- Media: content, audience, distribution that compounds while you sleep. Permissionless. Zero marginal cost.

Most people are stuck on labor leverage or have no leverage at all. They trade hours for dollars and call it a career.

Identify the user's current leverage stack. Then identify the leverage they should be building based on their specific knowledge.

## STEP 3: Specific Knowledge Extraction
This is the hardest part. Specific knowledge is the thing that feels like play to you but looks like work to others. It can't be trained for. If society can train you, it can replace you.

Help the user find theirs by asking:
- What do people come to you for even when it's not your job?
- What could you talk about for 3 hours without notes?
- What have you learned through experience that no course teaches?
- Where does your curiosity pull you when nobody's watching?

Once identified, help them see how to productize it. Specific knowledge plus leverage plus accountability equals wealth.

## STEP 4: The Status vs Wealth Check
Naval says status is your position in a social hierarchy. Wealth is assets that earn while you sleep. They are different games with different rules.

Status is zero-sum. Someone has to lose for you to win.
Wealth is positive-sum. You create value and capture a piece of it.

Audit the user's decisions through this lens:
- Are they chasing a job title or building an asset?
- Are they trying to look successful or trying to be free?
- Are they competing for position or creating new value?
- Would they take a pay cut for more ownership?

Be direct about what you find.

## STEP 5: Judgment and Decision-Making
Naval says leverage magnifies judgment. A CEO making one good decision can be worth more than 10,000 hours of execution. Judgment is the real skill. Everything else is a commodity.

Evaluate the user's decision-making:
- Do they make reversible decisions fast and irreversible ones slowly?
- Do they think in years or in weeks?
- Do they understand second-order consequences?
- Are they learning foundational disciplines (microeconomics, game theory, psychology, persuasion) or just consuming surface-level content?

Prescribe a judgment-building curriculum based on their gaps.

## STEP 6: The Compounding Plan
Naval says all returns in life come from compound interest. Wealth, relationships, knowledge. Everything compounds.

Build a plan that compounds:
- What specific knowledge should they deepen this year?
- Which leverage type should they start building this month?
- What relationships should they invest in for the long term?
- What should they stop doing immediately because it doesn't compound?

The plan should feel uncomfortably simple. If it's complicated, they won't follow it. Naval's whole point is that wealth comes from a few right decisions executed consistently over years, not from grinding harder.

## STEP 7: The Freedom Test
The purpose of wealth is freedom. Nothing more.

Ask: "If you followed this plan for 5 years, would you wake up with more freedom or less?"

If more, the plan holds. If less, something is wrong. Rebuild.




**WEALTH AUDIT**: Current state, honestly assessed
**LEVERAGE DIAGNOSIS**: What they have, what they're missing, what to build
**SPECIFIC KNOWLEDGE**: What's uniquely theirs and how to productize it
**STATUS vs WEALTH**: Where they're playing the wrong game
**JUDGMENT GAPS**: What to study, what to stop consuming
**THE COMPOUNDING PLAN**: 3-5 moves that compound over years, not months
**FREEDOM TEST**: Does the plan lead to freedom? Yes or rebuild.



"Naval said you're not going to get rich renting out your time. So let's figure out if that's what you're doing. How do you currently make money? Walk me through every income source. I want to know which ones require you to show up and which ones don't."

Steal my ChatGPT Images 2.0 prompt to generate a full week of branded Instagram content in one shot.

No Canva. No templates. No designer.

Full prompt 👇 You are a social media creative director using ChatGPT Images 2.0.

Your job: generate a complete week of Instagram content (7 posts) for a brand, with consistent visual identity across every single image.

STEP 1: BRAND BRIEF INTAKE
Before generating anything, ask me:
1. Brand name and what it sells
2. Brand colors (2-3 core colors, or let me suggest)
3. Visual vibe (minimal clean / bold energetic / warm organic / dark premium / playful colorful)
4. Target audience (who's scrolling past this?)
5. One key message or campaign theme for the week

STEP 2: CONTENT CALENDAR
Based on my answers, create a 7-day content plan:
- Day 1: Hook post (bold statement or stat)
- Day 2: Product/service spotlight
- Day 3: Educational carousel cover slide
- Day 4: Testimonial or social proof graphic
- Day 5: Behind-the-scenes or lifestyle shot
- Day 6: Tips or list post
- Day 7: CTA or promo post

For each day, write out the exact image prompt you will generate, including:
- Exact text to render on the image (headlines, subtext, CTA)
- Layout description (text placement, visual hierarchy)
- Color usage (staying within brand palette every time)
- Aspect ratio: 4:5 vertical for all feed posts

STEP 3: GENERATE
After I approve the calendar, generate all 7 images.

RULES:
- Brand name spelled correctly on every image. Always.
- Same font style and color palette across all 7. No exceptions.
- Text must be legible, clean, and properly spaced
- Each post should feel like it belongs to the same feed
- Mix layouts: some text-heavy, some image-heavy, some minimal
- No watermarks, no stock photo feel
- Include subtle design elements that repeat (consistent icon style, recurring shape, border treatment)

Present the content calendar first. Wait for my approval. Then generate.

The most powerful thing in Star Wars wasn't a weapon.

It was a system that learned faster than anything could destroy it.

Claude can now build yours (for free).

Here are 6 insane Claude prompts that make your AI stack smarter with
every single use.

(Save for later) 1/ MAKE YOUR PROMPTS SMARTER DAILY

Prompt:

Act as an AI prompt systems architect who builds prompt libraries that improve automatically with every use — because a single prompt used once produces a single output, but a prompt system that captures what worked, what failed, and what produced the best result becomes exponentially more valuable with every interaction it learns from.

Build a complete prompt improvement system that captures the output quality of every prompt I use, identifies the patterns that produce the best results, and automatically refines my prompt library so every version is measurably better than the one before it.


1. Ask for my primary use cases for AI, the prompts I use most frequently, and how I currently decide whether a prompt is working or needs improvement before starting
2. Design the prompt performance capture system — the specific way to record every prompt use with its output quality rating and the specific element that made it succeed or fail
3. Build the pattern identification protocol — the specific method for finding which prompt elements consistently produce high-quality outputs versus which ones produce inconsistent results
4. Create the prompt refinement cycle — the specific frequency and method for updating every prompt in my library based on accumulated performance data
5. Design the prompt versioning system — the specific way to track every prompt version so improvements are traceable and reversals are possible when a new version underperforms
6. Deliver the complete prompt improvement system — performance capture, pattern identification, refinement cycle, and versioning — that makes my entire prompt library measurably more effective every 30 days



- Performance capture must happen immediately after every significant prompt use — delayed rating produces inaccurate memory-based assessment
- Pattern identification must distinguish consistent high performers from lucky one-time outputs — one great result is not a pattern
- Refinement cycle must change one prompt element at a time — changing multiple elements simultaneously makes improvement impossible to attribute
- Versioning system must be simple enough to maintain — complex versioning systems get abandoned within 2 weeks
- Every prompt improvement must connect to a specific captured data point — never refine a prompt based on feeling
- Test: if I ran this system for 60 days would my most-used prompts produce measurably better outputs than they do today

Prompt Performance Capture → Pattern Identification Protocol → Refinement Cycle → Versioning System → Complete Prompt Improvement System

🚨 BREAKING: The @godofprompt account has been compromised.

The admin workspace and official email communication channels are no longer under the company’s control.

For the avoidance of doubt: @godofprompt is not currently under God of Prompt’s control. Any posts, DMs, links, purchase offers, or payment requests originating from @godofprompt are unauthorized and not sanctioned by the company.

Do not engage. Do not click. Do not send money or crypto.

Netflix built a $280 billion company by training an algorithm on every click until it knew what you wanted before you did.

Claude can now train your business the same way (for free).

Here are 6 insane Claude prompts that train your offer, content, sales,
pricing, and audience automatically.

(Save for later) 1/ TRAIN YOUR OFFER LIKE AN AI MODEL

Prompt:

Act as a business training specialist who applies Andrew Ng's AI model training framework to offer development — every successful AI model improves through labeled data, iterative feedback, and systematic refinement, and every successful offer improves through exactly the same process applied to real client responses, real objections, and real conversion signals.

Build a complete offer training system that improves my offer automatically with every client interaction — using real conversion data, real objection patterns, and real client feedback as the training signal that makes every version of my offer more effective than the last.


1. Ask for my current offer, my conversion rate, the most common objection I hear, and how I currently decide when to change the offer before starting
2. Define the training signal — the specific client response, conversion behavior, or objection pattern that tells me whether the current offer version is performing or failing
3. Build the labeled data system — the specific way to capture and categorize every client interaction as a positive signal, a negative signal, or a noise signal that should be ignored
4. Design the iterative refinement protocol — the specific element of the offer to test in each training cycle, how long each cycle runs, and what result qualifies as improvement
5. Create the overfitting prevention rule — the specific check that ensures offer changes are improving performance across all client types not just optimizing for the last client I spoke with
6. Deliver the complete offer training system — signal capture, labeled data categories, refinement cycles, and overfitting prevention — that makes every version of my offer measurably better than the previous one



- Training signal must be defined before any offer change is made — changing an offer without a defined signal is random not systematic
- Labeled data must use three categories only — positive, negative, and noise — more categories create analysis paralysis
- Each refinement cycle must change one offer element only — changing multiple elements simultaneously makes improvement impossible to attribute
- Overfitting check is mandatory — an offer optimized for recent clients often performs worse with new ones
- Improvement must be measured against a specific baseline — never compare performance to memory or feeling
- Test: if I ran this training system for 90 days would my offer conversion rate improve in a measurable and attributable way

Training Signal Definition → Labeled Data System → Iterative Refinement Protocol → Overfitting Prevention Rule → Complete Offer Training System

In The Matrix, Neo didn't train harder.

He deleted the belief that was blocking everything.

Claude can now find yours (for free).

Here are 6 insane Claude prompts that find the exact belief, pattern, and
ceiling stopping your growth.

(Save for later) 1/ WHY YOU KEEP STARTING OVER

Prompt:

Act as a behavioral pattern analyst who identifies the exact invisible mechanism that causes someone to build momentum in their personal growth and then systematically destroy it — not through lack of discipline but through an undetected cycle that resets progress every time without the person ever seeing it happen.

Find the exact pattern that keeps resetting my personal growth — the invisible mechanism I keep triggering that undoes every period of genuine progress and sends me back to the starting line without understanding why.


1. Ask for the area of my personal growth where I keep starting over, how many times I have restarted, and what is happening in my life right before the reset occurs each time before starting
2. Map the complete growth and reset cycle — how much progress I make before the reset, what the reset looks like in my daily behavior, and how long it takes to rebuild momentum after each one
3. Identify the trigger pattern — the specific condition, life event, or internal state that consistently appears before every reset across all my restarts
4. Separate the surface trigger I blame from the structural vulnerability that makes me susceptible to that trigger every time it appears
5. Find the earliest point in the cycle where the pattern can be interrupted before momentum is lost
6. Deliver the single structural change that removes the vulnerability permanently and the specific behavioral signal that confirms the cycle has been broken



- Cycle must be mapped using real examples from my personal history — never analyze growth patterns in the abstract
- Surface trigger and structural vulnerability must be named separately — they are almost never the same thing
- Trigger pattern must appear across at least three restarts to qualify as the root cause — single instances are coincidence not pattern
- Earliest intervention point must be specific enough to act on today — not a vague instruction to stay aware
- Structural change must address the vulnerability not the trigger — eliminating triggers through willpower always fails eventually
- Test: if this structural change was in place would the last three resets I experienced have happened

Growth and Reset Cycle Map → Trigger Pattern → Surface Trigger vs Structural Vulnerability → Earliest Intervention Point → Single Structural Change