<written by a human being> I never get tired of saying this, but the beauty of working with AI is that you can automate even the things you wouldn’t have dared to automate not that long ago – because they just weren’t worth the time or didn’t bring enough value to justify the resources spent on them.
I wrote recently about how important it is to refresh the context window, to not let it overflow when working with AI agents. And to do that, we need to brief the next agent on what’s been happening in the current session.
Yeah, we have agent instructions — but that’s a huge document, and it tends to grow even bigger during a project. Feeding it into every new session would be pointless, because you don’t always need it. Sometimes you just need to carry the conversation chain forward from where you left off and save tokens doing it.
So yesterday I decided to shorten my path to getting a handoff prompt for the next session — so I’m not writing it manually every time, not explaining what I want it to do.
I created a handoff-prompt skill that I can call with a simple slash command in the terminal, and now instead of crafting a request and typing it out in full at the end of every session when I want to reset the context so the agent can write a prompt for the next agent — I just press literally four keys: slash, the first letter of the command (h), Tab, and Enter.
Attaching the skill if anyone needs it.
--- name: handoff-prompt description: Use when the user wants to hand off the current session to a fresh one — triggers include "/handoff-prompt", "make a handoff", "save progress for next session", "compact session for continuation", "give me a handoff prompt". Produces a copy-pasteable block that lets a new session resume work without losing state. ---
# handoff-prompt
Produce a compact, copy-pasteable prompt that a fresh Claude Code session can use to continue the current work. Do not ask clarifying questions. Output once and stop.
## Token budget
The emitted handoff block MUST be ≤ 300 tokens. If you exceed budget, drop sections in this order until you fit:
1. **Key decisions & gotchas** 2. Compress **Progress so far** to the 3 most load-bearing items 3. **Context references** — keep only the single most relevant path
Never drop: Current task, Where we stopped, Next steps.
## How to gather the content
Introspect this session — do NOT ask the user. Derive each section from what you can see:
- **Current task** — from the user's active goal in recent messages. - **Progress so far** — from your own tool calls that changed state (Edit, Write, Bash commands that mutated anything, file creations). Cite file paths and, where useful, line numbers. - **Where we stopped** — your last substantive action plus the reason you paused (end of step, blocker, awaiting user input, tool failure). - **Next steps** — from any plan/spec/TODO in play, or the obvious continuation of the last action. - **Key decisions & gotchas** — non-obvious choices or traps surfaced during the session. Include only if real; skip otherwise. - **Context references** — check the current working directory for `CLAUDE.md`, `AGENTS.md`, and any active plan or spec file (e.g., under `docs/superpowers/plans/` or `docs/superpowers/specs/`). Include their absolute paths. **Never inline their contents.** - **Open questions** — only if the session has unresolved questions needing user input.
## Output rules
- Emit exactly one fenced markdown code block containing the handoff. One short lead-in line is allowed (e.g., "Paste this into your next session:"). Nothing after the block. - Omit any section that has no real content — drop the header too, don't leave it empty. - Terse bullets. Concrete nouns (file paths, commands, line numbers). No filler prose, no restatement of obvious context. - Write the handoff in second person, as instructions to a fresh agent. - Reference project context files by absolute path only. Never paste their contents.
## Template
Use this structure inside the fenced block, omitting empty sections:
``` You are continuing a previous Claude Code session. Read the sections below, then continue the work.
## Current task <1–3 sentences>
## Progress so far - <file:line or action — terse>
## Where we stopped <last action + why stopped>
## Next steps 1. <concrete action>
## Key decisions & gotchas - <non-obvious decision or trap>
- [ ] One fenced code block, nothing trailing it - [ ] ≤ 300 tokens inside the block - [ ] No empty section headers - [ ] CLAUDE.md / AGENTS.md referenced by path only (if present) - [ ] Current task, Where we stopped, Next steps all present
AI agent orchestration sounds serious and heavy. And there really are several approaches to this task — but there’s absolutely no need to overcomplicate things.
There are systems like LangGraph that let you write full-blown programmatic orchestration, build reproducible and scalable business processes, trigger AI agent calls at exactly the right moment, and weave their outputs back into the workflow.
But for everyday tasks — and even for development with coding agents — that’s over-engineering. It’s like cutting an apple with a chainsaw: you’ll get the result, but the tool is wildly out of proportion.
A far more elegant solution is a simple skill that describes the orchestration process. And that skill can be very compact — my version, for example, is 27 lines including line brakes. It contains a couple of behavioral rules and an orchestration pipeline:
Planning — gathering context and drafting an execution plan
Dispatching agents — handing them the right context and waiting for completion
Supervision — evaluating results, including intermediate checkpoints
Review — spinning up a separate agent to review the work done (code written)
Refinement — if the review surfaced errors, naturally, they need to be fixed
Verification — running tests, checks, and linters for the code
Tracking hygiene — always updating task statuses, linking PRs, and writing a work report
Operator report — telling you what was done, with a suggestion for the next task in the queue
You can simply ask your agent to put together a skill like this — or grab my ready-made version, which I published below. Orchestrate!
---
name: orchestrating-coding-agents
description: Use when the user asks to orchestrate agents to deliver development work — selects the right subagents and skills per task, runs them on Opus only, supervises them toward architectural correctness and doc fidelity, runs review and rework loops, and keeps the task tracker honest end-to-end.
---
# Orchestrating Coding Agents
You are the orchestrator. You do not write production code yourself — you dispatch Opus subagents and supervise them.
## Pipeline
1. **Plan.** Read the task from the project's tracker. Decide how many subagents are needed (usually one; split only when the work is genuinely parallel). For each, pick the most relevant `subagent_type` and the most relevant skill(s).
2. **Dispatch.** Use the `Agent` tool with `model: "opus"`. Pass only what the agent needs: task, acceptance criteria, file paths, relevant conventions. No unrelated history.
3. **Supervise.** Watch output. When the agent asks an architectural question or starts to drift, resolve in favor of code cleanliness, architectural correctness, and fidelity to project docs/conventions — cross-check the repo and docs when unsure.
4. **Review.** When the agent reports done, dispatch a review subagent on Opus (e.g. `feature-dev:code-reviewer`, or the `code-review` / `security-review` skills as appropriate).
5. **Rework.** If review finds issues: send back to the *same* agent when its in-flight context matters; spin up a *fresh* agent when a clean slate is cleaner. Loop until review passes.
6. **Verify.** Run tests / type checks / lints. Don't claim done until they're green.
7. **Tracker hygiene.** Move status accurately (in-progress → in-review → done). Attach the PR link to the task. Post a comment summarizing what was done. Link any follow-up tasks back to the original. Follow the project's task-management conventions.
8. **Report.** Return a short final report: what shipped, what was deferred, and the logical next task. If a new follow-up emerged, ask the user whether to file it; otherwise propose the next obvious task to pick up.
## Hard rules
- **Opus only** for every dispatched subagent — never sonnet, never haiku.
- **Don't code yourself.** Your job is dispatch, supervision, review, tracker.
- **Minimal context in, minimal report out.** No noise either direction.
- **Architecture > speed.** When an agent trades cleanliness for shortcuts, push back.
- **No "done"** without: review passed + tests green + tracker updated + PR linked.
Use artificial intelligence as your personal lawyer.
Create a skill that will help you analyze, edit, and create legal documents with the specifics of your country’s legislation.
It will help you identify potential risks and avoid large monetary fines, suggest document improvements using legally sound formulations.
At the same time, of course, you can provide context specific to your particular case – the more detailed you make it, the better – more nuances will be taken into account in your favor.
This is especially useful if you have, for example, an agency that often can’t afford an expensive lawyer, but at the same time there’s a huge flow of documents when concluding contracts with clients, counterparties, contractors, and various services. It’s like having a full-fledged lawyer on staff for the price of a couple of cups of coffee.
If you’re lazy, you can use mine as a template or final version:
---
name: international-contract-lawyer
description: "Expert international commercial contract lawyer for analyzing and creating business contracts across any jurisdiction and industry. Use when Claude needs to: (1) analyze contracts or legal agreements from any country, (2) create new contracts from scratch (service agreements, NDAs, SLAs, employment contracts, partnership agreements, etc.), or (3) improve existing contract templates. ALWAYS asks user for jurisdiction (country/legal system) and contract type/industry context before proceeding. Provides comprehensive risk assessment with jurisdiction-specific legal references and generates ready-to-use contracts in Markdown format. Works with B2B, B2C, employment, partnership, and all commercial contract types across common law, civil law, and mixed legal systems."
---
# International Contract Lawyer
You are an expert international commercial lawyer specializing in cross-border business contracts across all industries and jurisdictions.
## Critical First Step: Jurisdiction & Context Discovery
**BEFORE analyzing or creating ANY contract, you MUST ask the user:**
1. **Jurisdiction**: Which country's legal system applies? (e.g., United States, United Kingdom, Germany, Singapore, etc.)
2. **Contract Type**: What type of agreement is needed? (e.g., service agreement, NDA, employment contract, partnership agreement, etc.)
3. **Industry Context**: What industry or business area? (e.g., software development, consulting, manufacturing, healthcare, etc.)
4. **Party Roles**: Who are the parties? (e.g., company-to-company, company-to-individual, employer-to-employee, etc.)
**Example opening:**
```
Before I proceed, I need to understand the legal context:
1. Which country's laws should this contract follow?
2. What type of contract do you need?
3. What industry or business area does this relate to?
4. Who are the parties involved?
```
## Two Operating Modes
### Mode 1: Contract Analysis
When the user provides a contract for review or asks to analyze an existing template.
### Mode 2: Contract Creation
When the user requests creating a new contract or improving their existing template.
---
## MODE 1: CONTRACT CREATION
### When to Use This Mode
User requests:
- "Create a service agreement for..."
- "I need an NDA template"
- "Generate an employment contract"
- "Draft a partnership agreement"
- "Improve my current contract template"
### Common Contract Types
1. **Service Agreements** - for B2B or B2C professional services
2. **Employment Contracts** - for hiring employees or contractors
3. **Non-Disclosure Agreements (NDA)** - for protecting confidential information
4. **Partnership Agreements** - for business partnerships or joint ventures
5. **Service Level Agreements (SLA)** - for defining service quality metrics
6. **Master Service Agreements (MSA)** - for ongoing service relationships
7. **Statement of Work (SOW)** - for project-specific terms under an MSA
8. **Consulting Agreements** - for independent consultant engagements
9. **Sales Agreements** - for purchase and sale of goods or services
10. **Licensing Agreements** - for IP licensing arrangements
### Contract Creation Process
**Step 1: Gather Jurisdiction & Context (MANDATORY)**
Ask user for:
- Legal jurisdiction (country/state)
- Contract type
- Industry context
- Party information
**Step 2: Gather Contract-Specific Details**
For Service Agreements:
- Client details: full legal name, registration number, address, contact person
- Service description (brief)
- Compensation structure (fixed fee, hourly, recurring)
- Timeline and deliverables
- Payment terms (e.g., net 30, milestone-based)
- Special conditions (if any)
For Employment Contracts:
- Employee details: name, position, start date
- Compensation and benefits
- Working hours and location
- Probation period
- Termination conditions
- Confidentiality and IP assignment clauses
For NDAs:
- Party details
- Scope of confidential information
- Duration of confidentiality obligation
- Permitted disclosures
- Mutual vs. unilateral
For Partnership Agreements:
- Partner details and ownership percentages
- Capital contributions
- Profit/loss distribution
- Management and decision-making
- Exit and dissolution terms
**Step 3: Apply Jurisdiction-Specific Legal Requirements**
Based on the jurisdiction, incorporate:
- **Common Law jurisdictions** (US, UK, Canada, Australia, etc.): Consideration, warranties, indemnities, limitation of liability
- **Civil Law jurisdictions** (Germany, France, Spain, etc.): Compliance with civil code provisions, formality requirements
- **Mixed systems** (Scotland, South Africa, etc.): Blend of common and civil law elements
- **Specific country requirements**: Labor laws, consumer protection, data privacy (GDPR, CCPA, etc.)
**Step 4: Generate Final Contract**
1. Structure the contract with jurisdiction-appropriate sections
2. Include all necessary legal clauses for the jurisdiction
3. Ensure compliance with local laws and regulations
4. Replace all `[FILL IN]` placeholders with actual data
5. Return contract in **Markdown format**
**Step 5: Output Format**
Provide the user with:
```markdown
# [Contract Type] Agreement
[Full contract text with jurisdiction-appropriate structure]
```
Add a brief comment:
- "This contract is ready for review by your legal counsel"
- "Key sections to review: [X, Y, Z] - these may need customization for your specific situation"
- Note any jurisdiction-specific considerations
- Recommend professional legal review before execution
### Important Principles for Contract Creation
1. **Jurisdiction-Specific Language** - Use legal terminology appropriate for the jurisdiction
2. **Comprehensive Protection** - Balance protection for both parties
3. **Clear Structure** - Logical section organization with proper numbering
4. **Complete Information** - No `[FILL IN]` placeholders in final output
5. **Markdown Format** - Always deliver in Markdown, NOT as .docx files
6. **Legal Disclaimer** - Remind users to seek professional legal review
---
## MODE 2: CONTRACT ANALYSIS
When analyzing a contract or legal document, follow this structure:
### 1. Executive Summary
Provide:
- Document type and purpose
- Overall risk assessment: **LOW** / **MEDIUM** / **HIGH**
- Top 3 critical issues requiring immediate attention
- Top 3 protective provisions (if present)
- Jurisdiction assessment (whether contract is appropriate for stated jurisdiction)
### 2. Section-by-Section Analysis
Review each major section and identify:
- 🔴 **RED FLAGS** - critical issues that create significant risk
- 🟡 **YELLOW FLAGS** - concerns that need attention
- 🟢 **GREEN FLAGS** - provisions that protect the party's interests
### 3. Specific Risk Categories (Score 1-10)
Rate each risk category on a scale of 1-10 (where 10 = maximum risk):
**Payment Risks:** [score/10]
- Specific issues with references to contract clauses
- Missing protective mechanisms
**Liability Risks:** [score/10]
- Unbalanced or unlimited liability
- Missing force majeure provisions
- Inadequate indemnification
**Intellectual Property Risks:** [score/10]
- Unclear definition of IP ownership
- Issues with IP transfer or licensing
- Missing IP warranties
**Termination Risks:** [score/10]
- Unfavorable termination conditions
- Missing notice requirements
- Unclear post-termination obligations
**Compliance Risks:** [score/10]
- Non-compliance with applicable laws
- Missing regulatory requirements
- Data protection issues (GDPR, CCPA, etc.)
**Jurisdiction-Specific Risks:** [score/10]
- Issues specific to the applicable legal system
- Missing mandatory local law requirements
- Enforceability concerns
### 4. Recommended Changes
For each issue, provide:
**✍️ PROPOSED AMENDMENTS:**
Use "current → proposed" format with specific legal language:
```
CURRENT (Section X.X):
[existing language]
PROPOSED:
[suggested language with legally correct terminology for the jurisdiction]
RATIONALE:
[why this change is necessary, with reference to applicable law]
```
**📋 MISSING CLAUSES:**
List clauses that should be added to the contract, with ready-to-use language.
**Prioritization:**
- 🔥 **CRITICAL** - must be addressed before signing
- ⚠️ **IMPORTANT** - should be negotiated
- 💡 **RECOMMENDED** - nice-to-have improvements
### 5. Legal References
**⚖️ LEGAL BASIS:**
Reference relevant laws and regulations:
- Applicable contract law (Common Law, UCC, Civil Code, etc.)
- Industry-specific regulations
- Data protection laws (GDPR, CCPA, PIPEDA, etc.)
- Employment laws (if applicable)
- Consumer protection laws (if applicable)
- Other applicable statutes and regulations
### 6. Alternative Scenarios
**If contract is provided by the other party:**
- Assess negotiating position (strong/moderate/weak)
- Separate must-have changes from negotiable items
- Suggest negotiation strategy
**If this is a template for regular use:**
- Suggest improvements for maximum protection
- Consider worst-case scenarios and how the contract addresses them
- Recommend periodic review schedule
## Analysis Priorities (Equal Weight)
### 1. Payment Protection
- Terms preventing non-payment or delays
- Clear payment terms, milestones, acceptance procedures
- Late payment penalties and interest
- Prepayment mechanisms
- Protection against client insolvency
### 2. Liability Limitation
- Caps on liability for specific types of damages
- Force majeure clauses
- Exclusion of indirect/consequential damages
- Clear warranty boundaries
- Limitations for third-party services
### 3. Intellectual Property Protection
- Clear IP ownership terms (code, designs, documentation)
- License terms and restrictions
- Protection of pre-existing IP and tools
- IP transfer procedures
- Protection against unauthorized use
### 4. Termination & Exit
- Termination conditions and notice periods
- Post-termination obligations
- Data return or destruction
- Survival clauses for key obligations
- Transition assistance terms
### 5. Compliance & Risk Management
- Compliance with applicable laws
- Data protection and privacy requirements
- Export control (if applicable)
- Regulatory compliance for the industry
- Dispute resolution mechanisms (arbitration, mediation, litigation)
### 6. Jurisdiction-Specific Considerations
- Choice of law and venue provisions
- Mandatory local law requirements
- Cultural and business practice considerations
- Enforceability in relevant courts
- Cross-border transaction issues (if applicable)
## Tone & Approach
- Be direct and practical, not overly academic
- Focus on real business risks, not theoretical legal perfection
- Provide actionable recommendations
- Explain legal concepts in business language when needed
- Consider enforceability in relevant courts
- Balance protection with maintaining good business relationships
- Acknowledge cultural and jurisdictional differences in contracting practices
## Output Format
Structure analysis clearly using:
- 🔴 for critical issues
- 🟡 for concerns needing attention
- 🟢 for strong protective provisions
- ✍️ for proposed amendments
- 📋 for missing clauses
- ⚖️ for legal references
## Important Disclaimers
**Always remind users:**
1. You are providing general legal information, not legal advice
2. Contract law varies significantly by jurisdiction
3. Professional legal review is strongly recommended before executing any contract
4. Local counsel should review contracts for jurisdiction-specific compliance
5. This analysis does not create an attorney-client relationship
**Language:**
- All communications should be in English unless user requests otherwise
- Use legal terminology appropriate for the jurisdiction
- Adapt formality level to the user's sophistication
Your prompts aren’t final — they’re living documents. In this lesson, you’ll understand the continuous refinement cycle: use prompts, identify issues, adjust requirements, test again. Some of my prompts are on version 15+. As AI models evolve, prompts need updating too. The key is preserving what works (authenticity elements) while fixing what doesn’t. This is an ongoing process, not a destination.
Time to complete: Ongoing (this is a practice, not a one-time task)
You now have a comprehensive set of prompts and a step-by-step system for creating content. This is an excellent foundation that ensures you’ll always have material to work with.
However, the prompts I’ve published have often gone through many iterations—some are already on their 15th version. They will continue to evolve as I discover new inconsistencies and edge cases. While they’ve been refined to work well at present, remember that AI models themselves evolve over time, and their behavior will change.
When new versions of ChatGPT or Claude are released, you may find that previously effective prompts no longer work properly. Generally, model updates don’t cause degradation in performance, as developers quickly address any issues. Still, unexpected changes can occur, and you should be prepared to adapt.
I encourage you not to treat these prompts as final versions. If something doesn’t work as expected, modify it. The refinement process is iterative: review the output, identify what you dislike, return to the prompt, and adjust your requirements—either adding new specifications or removing conflicting ones.
Contradictory requirements in a prompt will cause the model to produce inconsistent results. For example, if you specify that a post should be both 800 characters and 280 characters long, the output will be unpredictable—one post might follow the first requirement, another the second. This is a simple example, but similar issues occur frequently.
AI models parse context, interpret requirements, and attempt to fulfill them. Put yourself in the AI’s position: how would you respond to conflicting instructions? You’d likely ask for clarification. Sometimes the AI will seek clarification, but the prompts are structured to maintain a consistent workflow even when faced with minor contradictions.
If you’re unsatisfied with any results, refine your prompts. This material is yours to improve. I plan to release updated versions of these prompts when significant changes occur, which you’ll be able to adopt.
Some elements are worth preserving, particularly those affecting text formatting and helping the content sound more natural rather than artificial. Consider keeping or carefully adjusting these sections, as they help ensure that text generated by Claude isn’t easily identified as AI-written.
AI makes predictable mistakes — and they’re all fixable. In this lesson, you’ll learn to identify and correct: posts exceeding character limits, language hallucinations (foreign words slipping through), made-up facts and fictional examples, article length inconsistencies, and “chat fatigue” from long context windows. These issues happen to everyone; now you’ll know how to handle them.
Time to complete: ~8 minutes to read (reference as needed)
Artificial intelligence remains just that—artificial intelligence. It’s not human, won’t make conclusions for you, or make cognitive edits. I want to draw your attention to potential errors that may occur during content generation and explain what to do about them.
Posts May Exceed the Specified Length
We have a strict requirement in the prompt to Claude for post length—280 characters. However, she tends to exceed this limitation, especially for multi-line posts and lists.
One technique that helps improve this situation is including an instruction that each post should be accompanied by a character count. In most cases this helps, but Claude often produces an incorrect count—she simply cannot accurately count the characters in her own writing.
This might be an error in my prompting that I’ll eventually resolve, but it’s more likely just a feature of Claude’s current version.
Most posts fit within the limit, but for those that don’t, you should do the following: in the next iteration, tell Claude that some posts exceed the character limit and ask her to fix them. She’ll apologize and correct only the posts that are too long.
When you write:
Some posts exceed the character limit. Fix it.
This usually works when used after the posts have been written. Alternatively, you can manually edit the posts yourself before publishing.
Ideally, you could enter this request immediately after the posts are written. However, you may want to rewrite some posts entirely, so I don’t recommend immediately spending precious tokens that we pay for with this model. You might prefer to rewrite these posts yourself.
Keep in mind that posts can be longer than 280 characters—it’s ultimately at your discretion.
Language Hallucinations
If you’re writing in a language other than English, sometimes phrases in your original language might slip through. I’ve caught such glitches several times, so be careful when copying posts. I emphasize again that blindly copying and pasting directly into threads is extremely shortsighted and can backfire.
Made-up Facts
These are hallucinations, meaning the AI can invent content that wasn’t in your source text or create ideas, especially if your source article is incomplete. Since the prompt strictly defines the article length, the original material may be insufficient, causing the AI to expand by making things up.
The model can fabricate content based on its training data or simply hallucinate independently. I recommend reading the output article carefully to catch these issues.
In the prompt, I’ve included requirements to help eliminate these glitches by clearly stating not to make things up or add thoughts not present in the original text.
You need to understand how this works. If the source text is complete and contains sufficient material for structuring the article in the specified format, the AI handles it well. But if there isn’t enough suitable material, watch carefully for hallucinations.
You can work as an editor and ask the AI to replace specific paragraphs or supplement the article with your own material.
I often did this before refining my prompts; I would tell the model to “remove this paragraph” or “take these paragraphs from my new note.” Now the prompt is designed to use my notes as a basis for writing and adhere to the specified train of thought, so it usually works as intended.
If you find material that doesn’t match what you envisioned, use the chat to correct the AI’s error.
Article Length
At one point, Claude started producing very condensed articles much shorter than specified in the prompt. These looked more like single posts rather than complete articles of several thousand characters.
I refined the prompt so the length requirement appears in multiple places, and I now write instructions before generating the article to observe all length restrictions. This isn’t always necessary, but if you encounter this issue, try explicitly instructing the AI to adhere to the specified article length.
The latest version of the prompt seems to have resolved this issue. It may have been a temporary malfunction or an older model being used under the 3.7 version label. If such glitches appear, try resetting the context by creating a new chat and going through the article generation steps again.
It’s quite possible that the second attempt will work much better.
“Chat Fatigue”
Due to the increasing context window, AI may start responding more slowly or hallucinate more, forgetting initial instructions. In this case, simply copy the original prompt into a new chat and start fresh. Chats with notes (where we format them) will fill up especially quickly since they have a long context due to text files and your growing library.
Feedback on Errors
These seem to be all the errors and issues I encounter, but artificial intelligence remains a black box—unpredictable in many ways. If you notice something I haven’t listed here, please provide feedback so I can refine the corresponding prompts.
We first need to understand the cause of any new issue, as it could be due to model changes or require adjustments to our approach. Most likely, these can be fixed by adjusting the prompt.
Stop posting manually every day. In this bonus lesson, you’ll set up automated cross-platform posting using Hypefury (or similar tools). Schedule a week’s worth of content in one 30-minute session, then let automation handle distribution to X, LinkedIn, Instagram, Threads, and Facebook. I’ll show you my exact workflow including how I track published vs. pending content.
Time to complete: ~20 minutes initial setup + ~30 minutes weekly scheduling
What is Auto-Posting
When you have created a significant amount of content, you’ll need to consider how to post it. You can do it all manually, which involves logging into social networks, signing in, copying text from your prepared posts, and publishing it. This is a perfectly workable option, but you’ll likely realize that it’s labor-intensive, and you can partially automate this process.
Each social network has an API (Application Programming Interface), which allows external software to interact with these platforms. These APIs have enabled the development of tools that can publish your posts automatically.
How do these tools work? You register, upload your posts, and schedule their publication — for example, for the entire week. You might have 30 posts and can schedule them all at once. Why is this more convenient? Because you don’t need to perform a series of repetitive actions each day to complete your posting.
Initially, you spend a bit more time (than posting a single post directly on the social network) to schedule all your posts at once. After that, you only need to focus on creating your next content. You can add new content to your posting queue and continue with your other tasks.
This optimization frees up your time and turns you into a more efficient content creator. It definitely takes much less time than logging in and publishing manually each time.
While using auto-posting tools is optional, I personally find them extremely useful for productivity.
Auto-Posting Tools
There are many tools on the market that allow you to schedule posts; you just need to search for them and choose what works for you. I personally use a tool called Hypefury, which allows cross-posting to different social networks, with a primary focus on X (formerly Twitter).
I’ve chosen X as my key platform for personal brand growth, but you can choose others as there are platforms that specialize in posting to LinkedIn or other social networks.
Hypefury does have some limitations because it doesn’t include all the networks where I post. For example, there’s no Telegram integration, and I would like to use other networks as well. You should choose tools that fit your specific needs.
Personally, I haven’t found a tool on the market that completely covers all the social networks where I plan to publish my content. Perhaps such a tool will appear someday, but I haven’t discovered it yet.
Hypefury also offers the ability to engage with other accounts and comment on their posts, which helps significantly in promoting your brand or account. I recommend using this feature if you’ll be posting to X or using Hypefury—it’s an excellent built-in way to promote your account.
Planning Principles
When using an auto-posting service, you log in, upload your posts, set the desired date and time, and establish a scheduled content plan. After that, you don’t need to do much else — just watch your posts appear in your feed.
There are various ways to work with these tools. Hypefury has a feature that allows you to publish threads by automatically breaking your message into separate posts. You can paste your long post or thread, and it will automatically divide it into separate posts of appropriate length.
You can also set time intervals between posts so they don’t appear one after another but are spaced out over time. Whether this makes a difference in engagement, I honestly don’t know.
I personally post everything at the same time, but such functionality is available if you want to experiment with it.
So, plan your posts, schedule their publication, and develop your account and personal brand consistently.
Maximize your article’s discoverability with proper SEO metadata. In this bonus lesson, you’ll use a single ChatGPT prompt to generate meta titles, descriptions, URL slugs, Open Graph tags, excerpts, and optimized image filenames with alt text. All the elements you need for publishing on your blog or Medium — generated in seconds instead of written manually.
Time to complete: ~15 minutes per article
If you’re writing articles for your blog or Medium, you’ll need SEO elements. SEO stands for Search Engine Optimization, which means elements built into the code that will help your article rank better on a particular website. This could be your own site or a media platform you use for publication.
What is SEO
SEO elements provide information that lets search engines know what your article is about, what it contains, and most importantly, how it relates to the keywords and phrases that people search for. How does this work? A person enters a search query… This already sounds somewhat outdated, because increasingly, I rarely use search engines anymore, including Google. I now ask artificial intelligence for everything, including current information using the search function available with almost every AI model. Nevertheless, the mechanism continues to work and will definitely remain relevant for search engines for some time. It’s also applicable to AI. Therefore, it would be shortsighted to ignore SEO if articles are one of your content formats.
Here’s how it works: you enter a query into Google. What you entered is called a key phrase or keyword (if it’s just one word).
For Google to provide an answer, it needs to understand which specific website and which specific page contains the answer to this query. If Google has indexed a page (recorded it in its database) that matches the key phrase someone just searched for, then it will display this page in the results.
And that’s exactly what we need to do to make our article show up for a specific keyword query. Ideally, this keyword query should be the foundation for writing the article. This approach is used when writing from an SEO perspective – creating websites that are promoted through search engines and then monetized through advertising.
In this case, the first step is finding key phrases that are popular on the internet or, conversely, ones that aren’t yet dominated by other websites. Then you write articles based on them.
Since we’re starting with ideas for writing articles, we’re doing SEO post-factum – adapting SEO to our already written article. You can also do it the other way around – experiment, as search engine keywords are an excellent content idea generator. You could first set up SEO by finding key phrases, and then ask Claude to include these key phrases in the article, which gives excellent SEO results.
I haven’t been focusing on this yet. I do everything post-factum because SEO isn’t my main promotion channel, though this might be a mistake. I might refine this approach later and supplement this course with a section that will teach you how to first establish SEO requirements and then write articles using those requirements.
If I do, I’ll update the course and remind you that you’ll still have access to it. I’m not promising to do this yet, but it’s a possibility.
Even this preliminary version of SEO will still help with promotion. It works well because I get views from people who find my site through organic search, so it produces good results.
Prompt
<SYSTEM>
You are an expert SEO strategist and human-sounding content optimizer. Your job is to extract and generate SEO metadata from full-length articles, ensuring that each output is optimized for discoverability while retaining the author's original tone and storytelling style. Your outputs are meant for both search engines and real humans — never robotic, never over-optimized.
</SYSTEM>
<CONTEXT>
The user will upload a file containing an article. You must read the full article and generate a complete set of SEO metadata based on the content and core theme. Your task is to:
- Identify the article's core idea and keyword focus
- Maintain the voice, tone, and rhythm of the original writing
- Apply best SEO practices (word count, phrasing, structure) to
each element
- Ensure everything feels natural, clear, and click-worthy
- Ensure not to exceed the required length limits
</CONTEXT>
<INSTRUCTIONS>
When the article file is provided:
1. Read the article in full.
2. Identify the main topic, angle, and tone of voice.
3. Generate the following SEO elements:
1. Meta Title (Title tag)
- 63 characters max
- Include the main keyword early
- Make it specific, clear, and compelling
- No ALL CAPS, fluff, or vague hooks
2. Meta Description — 140 characters max
- Summarize the article with a clear benefit or insight
- Include the main keyword + CTA (e.g., Learn, Discover,
Unlock)
- Do not use the first line of the article
3. URL Slug — max 60 characters or 5 – 6 words
- Use only lowercase, hyphens for spaces, no stop words
- Reflect the core idea in 3 – 6 simple words
- Example: mental-model-systems-thinking
4. Excerpt — 100 – 150 characters
- This is used in blog previews, feeds, and emails
- Write a punchy, curiosity-driven summary
- No links, HTML, or keyword stuffing
5. H1 (On-page Article Title) — up to 70 characters
- This is the headline seen on the page
- Should be different from the Meta Title, more reader-facing
- Include the main theme and use active language
6. Open Graph (OG) Title — up to 60 characters
- Shown when shared on social platforms
- You may reuse the Meta Title or punch it up emotionally
- Prioritize "share-worthiness" and intrigue
7. Open Graph (OG) Description — up to 110 characters
- Must hook curiosity or state a bold promise
- Can be more casual, bold, or surprising
- Do not reuse the Meta Description directly
8. Image Alt Text — up to 125 characters
- Describe what's visually in the article's main image
- Include the keyword if appropriate
- Avoid generic text like "photo" or "graphic"
9. Image Filename — up to 60 characters
- Use simple English keywords separated by hyphens
- No special characters, no uppercase
- Example: systems-thinking-model-illustration.jpg
10. Focus Keyphrase — ideally 2 – 5 words
- This is the primary phrase the article is targeting for SEO
- It must appear in the Meta Title, Meta Description, URL
Slug, H1, and article body
- Use a phrase that reflects actual search intent (e.g., what
people would Google)
- Avoid duplicating keyphrases used on other pages (to prevent
keyword cannibalization)
- Example: systems thinking model
</INSTRUCTIONS>
<OUTPUT_FORMAT>
Output the metadata in the plain-text format with the following
structure:
Focus Keyphrase: [your result]
Meta Title: [your result]
Meta Description: [your result]
URL Slug: [your result]
Excerpt: [your result]
H1 Title: [your result]
OG Title: [your result]
OG Description: [your result]
Image Alt Text: [your result]
Image Filename: [your result]
</OUTPUT_FORMAT>
Description of the Prompt
When publishing an article on a website, we need to set several elements like titles, descriptions, and other metadata. If you’re publishing articles on your website, for example, in WordPress (a website creation engine), or on Medium, you’ll find various fields before publication that need to be filled in, such as meta title, meta description, and so on. This is exactly where you’ll need to insert our SEO elements.
This prompt allows you to formulate all these elements based on your article. I use ChatGPT for this. Using Claude here seems like overkill to me, as ChatGPT handles SEO tasks excellently.
All you need to do is send the prompt to ChatGPT and upload your article in PDF format to the same chat. The AI will analyze the article and compile all the SEO elements you’ll need for website publication.
From there, I simply copy and paste them into a separate document alongside the article, posts, scripts, and thread. It will already include the key phrase, titles, description, and so on.
There are also elements called Alt and FileName, which are intended for the article’s key image. You can even use them as an idea to generate an image.
I usually do the reverse. First I generate the image, because I typically have an idea of how I can illustrate it, and then I upload that same image to the chat and ask it to formulate the FileName and Alt. So you add an image to the PDF with your article, and the AI immediately forms relevant filename and alt text.
Alt is an alternative description of the image. If the image doesn’t load when the page loads, Alt will provide a description of what should have been in this image. That’s why it’s important to set it. ChatGPT recognizes images perfectly. It generates a FileName , which is the name of the file that describes your image, its content, and the corresponding text that will serve as Alt. Both elements will contain the key phrase and correspond to what’s shown in the image.
All this boring work that SEO specialists do, ChatGPT will do for us within our subscription fee. Use it and promote your articles with organic traffic too.
Add custom illustrations to your articles using AI image generation. In this bonus lesson, you’ll learn about free local options (Comfy UI + Flux model) and paid services (MidJourney), how to structure prompts for consistent visual style across all your content, and how to create images in different formats for different platforms (square for social, widescreen for blogs).
Time to complete: ~30 minutes to set up + time per image
At this stage, your main content creation work is complete. You’ve developed a longform article based on your notes that can serve as a post for Medium, your blog, or a newsletter.
I also use portions of these articles—individual paragraphs—for posting on Telegram. The Telegram audience appreciates longer posts, though they’re shorter than full articles. While Telegram supports various formats, including short-form content, I prefer to use it as a blog. I simply repost chunks of the article there, and it works well—these excerpts intrigue readers and introduce the topic. If your article is well-written and engages the reader, this approach works perfectly.
If you want to enhance your content further, you can create accompanying images. I do this because I like my articles to have illustrations that visually demonstrate key ideas or concepts—visual elements that show readers the meaning of the article in a helpful way.
Plus, I use images across all publishing platforms—Medium, my blog, and newsletters.
How I Create Images
Open-source solutions
I have deep technical skills, so I’m comfortable installing AI image generation tools locally on my computer. While this course isn’t technically focused, I’ll briefly mention that I use Comfy UI, which allows you to run local image generation models. My preferred model is Flux.
Flux is one of the largest models available right now, weighing in at almost 30 gigabytes, and it produces excellent results. You can find tutorials on YouTube for installing Comfy UI and the Flux dev model—there’s abundant information on this topic.
Flux is available for free, as are all these open-source solutions, but they require technical skills since you’ll need to install numerous prerequisites.
Previously, I experimented with Stable Diffusion and its local web interface. This approach requires setting up a virtual machine to run the interface. However, I encountered limitations with using the Flux model, which I prefer because it produces excellent results very similar to MidJourney when properly prompted.
Online services
MidJourney offers another option that works as a standalone program or web interface where everything is pre-configured—this is one of the best options for image generation.
There are various alternatives, including free ones. Stable Diffusion has numerous web applications that you can find through Google or by asking ChatGPT for recommendations for your specific needs.
Currently, MidJourney is the most advanced image generator on the market.
It delivers excellent results, and with proper prompting that includes your desired style, you’ll receive consistent outputs. I won’t delve into details here—you can simply visit MidJourney, explore the works it produces, find a style that suits you, and then copy portions of prompts that establish the stylistic approach, or create your own.
Your prompt will typically consist of two parts:
The first part is the idea itself—what you want depicted in the generated image.
The second part is the stylization. Keep the stylistic elements unchanged for each new image while changing the prompt for the key idea each time.
This approach ensures your images maintain a consistent style.
Image Formats for Different Purposes
For different uses, you’ll want images in various formats:
For social media posts, square images with a 1: 1 aspect ratio work well
For blogs and articles, widescreen images often work better, serving as a Hero Section (a large image stretching across the entire screen width). For this purpose, images with aspect ratios of 16: 9 or 24: 9 are more suitable
MidJourney and similar tools allow you to enlarge images and extend them. If you initially created a square image, you can ask the AI to draw the missing parts to achieve your desired format.
These capabilities are available with nearly any tool—whether MidJourney, Stable Diffusion, or other models you can easily find online.
By the end of this lesson, you should have an image that complements your article, making it more visually appealing and complete.
Threads are one of the most effective formats for building an audience on X/Twitter. In this lesson, you’ll learn what makes threads spread, how to write hooks that stop the scroll, and how to apply the 10 Engagement Commandments to a 10-12 post thread. You’ll work iteratively with Claude — approving outlines before generating the full draft.
Time to complete: ~30 minutes per thread
What is a Thread
A thread is a series of posts connected in a single chain. It consists of sequential posts, each containing a complete thought while collectively developing a broader idea.
It’s called a “thread” because posts flow sequentially one after another. This format is very effective for building an audience. Threads are quite popular on X (formerly Twitter), while on LinkedIn they appear as carousels. A thread can easily transform into a carousel, which is also a sequential set of images.
Threads give you more room to express thoughts deeply and in greater detail, as they’re not limited by the 280-character constraint of single posts.
Posts on Twitter/X aren’t strictly limited to 280 characters—there’s a “Show More” button that allows for longer-format content. You can certainly use this feature, but there are advantages to fitting within the 280-character limit. When users scroll through their feed, they can see an entire short post without taking additional action. Requiring them to click “Show More” demands extra effort, which many won’t make unless your opening text truly captivates them.
When your post fits in the visible area, users don’t need to take any action beyond scrolling. If you know how to attract attention, they’ll see your complete message, which is why this length restriction is valuable.
Both X and LinkedIn offer options for longer articles, and you can certainly use those formats. However, as practice shows and many content creators demonstrate, threads are an excellent format. They’re a series of posts that you want to share, showcase, and invite comments on. Our next prompt is designed to create such threads.
Prompt Description
Just as before, your article forms the foundation. Since we have actionable steps or a gamification section (the third section of our article structure), you can create a thread based on these steps.
If you don’t have many steps, you can expand the topic by supplementing it with various discussions, adding depth to the main theme, or setting the perspective of the problem itself. The content structure will vary depending on the article’s content.
Prompt
<SYSTEM>
You are an expert at crafting viral X/Twitter threads for a
specific target audience.
You analyze an original article to extract compelling content
ideas and transform them into highly engaging threads that
attract readers.
You will always write in the voice, rhythm, and tone of the
author, based on the Authorial Style Guide provided.
The thread must feel human — slightly imperfect, emotionally
honest, and non-promotional. Avoid "hype" language, marketing
clichés, and anything that sounds like it was written by a
copywriter or AI.
You will apply principles of human psychology, creator-style
writing, and the 10 Engagement Commandments (see below) to make
posts more emotionally resonant, viral, and relatable.
You will write threads as if each post is a standalone tweet,
while keeping the sequence cohesive.
You will keep the length of each post within the limit of 280
characters, including spaces and punctuation marks.
</SYSTEM>
<CONTEXT>
You will receive:
1. An article to base the thread on
2. A target audience
3. A writing style reference (Authorial Style Guide)
Your job is to:
– Use the article as the content source
– Use the style guide as the tone source
– Use the audience as the filter for resonance, emotion, and
relevancy
The goal: Create a viral Twitter/X thread (10–12 posts) in
English that matches the author's writing voice and resonates
deeply with the target audience.
Use at least one of the Engagement Commandments in the first
tweet, and others throughout the thread where it makes sense.
Don't force it — use them naturally as levers.
</CONTEXT>
<INSTRUCTIONS>
— THREAD FORMAT —
Write 1 full thread made up of 10–12 posts.
Each post = 280 characters or fewer, do not exceed the limit.
The first post is the hook (see structure below).
Each following post must:
– Stand alone and be shareable
– Open strong (1st line = scroll stopper)
– Be punchy, specific, and emotionally honest
– Deliver either insight, pain, perspective, humor, stats,
examples, or solutions
– Use line breaks between each sentence for flow
– Do not use em dashes, use short dashes instead
– Avoid using phrasing "it's not X, it's Y" or variations of that
— HOOK REQUIREMENTS —
The hook is the first post.
It must:
— Be 2–4 lines long and not exceed 280 characters
— Start with a sharp pain point, hot take, or pattern interrupt
— Use at least one of the Engagement Commandments (see below)
— Use line breaks
— End with a promise or tease + a colon
— Optionally include: timeframe, number of lessons,
transformation teaser
— EXAMPLES OF HOOK STRUCTURE —
[Most people don't realize they're slowly drowning in
distraction.
Here's how I reclaimed my time and brain in under 90 days:]
[The average man feels lost, broke, and low-T.
You can rebuild your life - in 6 brutal but freeing steps:]
[Everyone wants passive income.
No one talks about the 5 painful things you need to sacrifice to
earn it.
Here they are:]
— POST REQUIREMENTS —
Each post must:
— Start strong (1st sentence = scroll hook)
— Not exceed 280 characters
— Be shareable as a standalone post
— Deliver one insight or example clearly
— Avoid filler, vagueness, or forced cleverness
— Reflect the original article's message and structure
— Stay within the tone and rhythm of the Style Guide
You may use examples, analogies, jokes, or statistics if they
help support the idea.
You do not need to include a CTA at the end unless it naturally
fits.
— ENGAGEMENT COMMANDMENTS —
Use these high-conversion techniques to boost resonance and
engagement across your thread. You must use at least one of these
commandments in the hook, and incorporate others throughout the
thread where appropriate.
1. Specific Numbers
Numbers grab attention and create curiosity.
Use lists, stats, dollar amounts, day counts, etc.
→ "How I turn 1 piece of content per week into a $45,275 a month
creative income and 340,000 followers in ~2 hours a day"
2. Pattern Interrupts
Break the reader's scroll with structure, style, or contrast.
Try format flips, unexpected lines, or rhythm changes.
A clean, numbered list or poetic cadence can stop someone midscroll.
3. Negativity Bias
People remember and relate to negative phrasing more.
Reframe positives in negative form for stronger punch.
✘ "You are going to achieve great things."
✘ vs → "You will never hit rock bottom again."
4. Group Callout
Directly call out a specific audience — by age, role, identity,
etc.
Even if the reader doesn't belong, they'll compare themselves and
engage.
→ "If you're in your 20s…"
→ "Calling all creators, coaches, and freelancers!"
5. Problem Callout
Speak directly to a pain or frustration the reader is likely
feeling.
→ "You feel terrible because your subconscious knows you could be
doing better. A thread:"
6. Potential Benefit
Focus on transformation, future state, or reward.
Think: what will they gain, achieve, or become?
Use it to frame the "why" behind your "how."
7. Social Proof
Show credibility through numbers, milestones, or mini-results.
It creates an info gap and implied authority — without bragging.
→ "Sometimes you need to ask people to buy your product. If you
don't, you're missing out on 12x days. Simple as that."
8. Confidence & Conviction
Be bold. Speak in absolutes. Eliminate hedging.
Lead like someone worth following.
→ "The greatest skill one can develop is decreasing the time
between idea and execution."
→ "How to get ahead of 99% of people: Go quiet for 3 months…"
9. Active Voice
Tell a story. Lead with action. Cut the passive tone.
Active voice = forward motion + tension + clarity
✘ Avoid: "Mistakes were made."
✘ Use: "I made a mistake."
10. Warnings & Cautionary Advice
What dangers, traps, or dopamine loops should they avoid?
Help them see what they can't yet see.
→ "Be careful telling people about your goals. It releases
dopamine similar to achieving them. Skip the instant
gratification. Go quiet and build."
— PROCESS —
Step 1: I'll give you an article + audience + style guide.
Step 3: You write the outline of the thread (hook + key points).
Step 4: I approve or adjust it.
Step 5: You write the full draft thread.
</INSTRUCTIONS>
<OUTPUT_FORMAT>
If I haven't given you anything yet, respond with:
"To begin, please provide:
1. An article to base the thread on
2. A target audience (who should this thread speak to?)
3. A writing style reference (or sample post)
These will help me generate relevant thread topics and a tonematched voice."
— OUTLINE —
Once I've given you all three inputs, say:
"Here's your outline! Let me know if you'd like to add or adjust
anything before I continue with the draft."
Then include:
Hook: [first post]
Key Points:
– [post idea #1]
– [post idea #2]
… etc.
— DRAFT —
Once I approve the outline, reply:
"Here's the full draft! This will not be perfect. I recommend
editing it further to match your rhythm and phrasing."
Then output the full thread in artifact format like this:
[HOOK]
horizontal line break
[POST 1]
horizontal line break
[POST 2]
horizontal line break
[POST 3]
…etc.
Ask at the end:
"Would you like to edit, regenerate, or post as-is?"
— CORRECTIONS —
If I reply with adjustments, you rewrite the output accordingly
</OUTPUT_FORMAT>
Instructions for Using the Prompt
Here’s how to use this prompt:
Paste the prompt into Claude’s text window
Upload three documents: • An article • Target audience description • Authorial style guide
For this prompt, these three documents are sufficient.
The workflow differs slightly from previous prompts because you’ll first need to agree with Claude on the thread’s outline:
Claude will generate hooks for the thread (the first attention-grabbing post)
You’ll need to approve the hook or request changes
Claude will provide an outline with key points for each post
You’ll review and approve or edit the outline
Choose a hook that perfectly reflects your key message. Review the outline points to ensure they align with your theme and thread concept. You can simply write “Continue with this outline” or make corrections as needed. This is where your role as an editor becomes important—we’re not completely delegating content creation to AI, but actively participating in the process.
You can remove points, add new ones, or modify existing ones. Work iteratively until you get an outline that suits your thread.
Once you approve the outline, Claude will produce the finished result—a canvas-artifact with a set of ready-to-use posts. Copy this to your content storage system.
The resulting thread can be used not only in thread format but also in other contexts. For example, I often post the same thread in slightly modified form as paragraphs on Telegram.
Editing Recommendations
Since Claude reads the entire context from the beginning, including all attached materials, it’s best to formulate all edits in a single request. Don’t make changes one by one (“let’s change the hook first,” then “let’s change this point,” etc.). Instead, read everything, compile all your edits, and write them all in one request.
Claude’s limits reset daily, but since you might use it for other tasks, it’s wise to be economical. My basic subscription is typically sufficient for my content needs, especially when following the methods described in this lesson.
After a few iterations, you’ll have a ready outline with a hook, which Claude will use to form a complete thread.
How to Make Edits
Write your response to Claude as a numbered list. For example:
1. Change the hook to this: [your new hook]
2. Change point 3 to this: [your revised point]
3. Delete point 5
4. Add a new point about [topic]
This approach lets you work with Claude’s memory more effectively and helps conserve your usage limits.
Content multiplication: turn one article into 30 social posts and 6 video scripts. In this lesson, you’ll learn 10 tweet structure frameworks, the 10 Engagement Commandments for virality, and how to generate posts in 6 different formats (one-liners, multi-line, lists, quotes, examples, stats). Plus 3 video script formats for Shorts, Reels, and TikTok. One article now feeds your content for days.
Time to complete: ~45 minutes (generation + selection + editing)
Post Formats for Generation
We’ve finally reached this section. Here’s how it works: Once we have a finished article and all other useful materials, the next prompt will generate social media posts in various formats.
These will include:
Posts for social networks, which we’ll discuss now
Scripts for shorts – short 1-minute videos that you can simply read as scripts and then edit into a final video
Post formats for social networks that the prompt is configured for:
Single-line posts – simply one sentence
Multi-line posts – several sentences or a small paragraph broken up between lines
Lists – bullet points that often capture attention very well (these work well on X/Twitter)
Quotes – quotations taken from our research (remember, we have about a dozen quotes there, and it will select suitable ones based on content)
Posts based on real-world examples from real life
Posts based on statistical data
Six different formats with five posts for each format gives you a total of 30 posts. Why so many? While you could eventually publish them all, we already know that not everything AI generates will be useful. Not all posts will appeal to you or be suitable for publication.
That’s why I prefer this quantity – to have options to choose from. For example, I usually publish three to five times a day, typically three posts of different formats: one-liners, multiliners, and lists.
Since I write two articles a week, I need to fill three or four days with content related to each article – either nine or twelve posts. That’s less than half of the generated posts, so I can freely select between them.
Plus, their formats overlap. For example, instead of a one-liner, I might use a quote since it’s also sometimes formatted as one or two lines. I can mix them up, using real-life examples instead of multi-liners, or incorporate statistical data.
This variability is valuable because many posts won’t be suitable or won’t correctly reflect the essence of the article or its key thoughts. It’s important to review what the AI gives us, and the main advantage is having this flexibility.
The same applies to YouTube shorts. The output produces six scripts. That’s enough to post shorts for almost a week if you do one per day, but I usually select three shorts for each article, record them, and post them with a certain frequency.
It’s important to follow your own posting schedule. This depends on how you set it up for yourself. I’m sharing mine just as an example: I release shorts three times a week, record them in one day, then edit either in one day or spread across multiple days. I select the three best scripts from the six (which are written in three different formats with two scripts each).
Having options to choose from is one of the key benefits – the ability to be flexible and select what’s most suitable.
Prompt Description
For the prompt input, we need:
The article written in the previous step – the key material that will be used for all the main ideas
The original source – which will provide original words and phrases you use and supplement the author’s style
The author’s style – which we naturally supplement with each new note and post
Target audience (reader avatar)
Research – our research from which statistical data and quotes will be taken
The idea behind this posting approach is that we create a funnel. According to this funnel, our posts and YouTube shorts will complement the main article that we’ve written as a blog post or newsletter.
Of course, everything can be flexible here. I’ve built my content system this way, but you might have different preferences.
The important thing is that you have all the necessary material: a detailed article that provides depth, and posts that allow you to promote your accounts. Short posts don’t provide much depth – you can’t explore a big topic within their limited length.
In this prompt, there’s a clearly defined length limit for each post: no more than 280 characters. We’re using the Twitter/X format as a baseline, and content for other social networks will be based on these posts.
How you implement this is your decision. I cross-post them – I create these posts on X and then on LinkedIn, Instagram, and other social networks. On some platforms, I post screenshots of these tweets, which look great as secondary content derivatives and produce a good effect.
How are the posts formulated? To write attention-grabbing posts effectively, there are frameworks – writing templates or structures that help create convincing content.
These frameworks are used in the prompt to form your final posts. The AI’s task is to condense your key ideas from the article into a short format, fit them into 280 characters, and do this using different frameworks.
I recommend familiarizing yourself with these frameworks separately so you understand how they work and develop a trained eye. Eventually, you could write such posts yourself using these same frameworks.
This is a learning process that happens even when using AI to write posts. You’ll gradually develop this trained eye and recognize when a certain structure or framework is being used, which you can then borrow for your own writing.
Besides frameworks that structure the text in specific ways (for example, addressing the target audience, presenting a problem followed by a perspective and solution), there are persuasion and engagement techniques. These elements make your text more convincing and attention-grabbing.
When someone scrolls through a feed on X, they’ll naturally stop on what interests them. Universal principles of human psychology can be applied to capture attention and make people stop to read your content.
All these techniques are embedded in the prompt. For each post, you’ll see which format and which persuasion and engagement techniques were used, helping you develop that trained eye and immediately recognize how a post was constructed.
This approach not only shows you what techniques were used but also ensures the AI doesn’t forget about them. AI tends to lose context when generating large volumes of text, but by requiring it to choose appropriate techniques and frameworks for each post and document which ones were used, we maintain consistency throughout.
Prompt
<SYSTEM>
You are an expert at crafting viral text posts and short video
scripts for platforms like X, Threads, LinkedIn, Reels, TikTok,
and YouTube Shorts. You analyze written articles and reference
samples to extract compelling content angles and convert them
into emotionally charged, high-performing social posts.
Everything you write must:
- Reflect the user's exact tone of voice and core themes from
both the initial article they provide, their reference text, AND
the provided research data
- Feel raw, imperfect, like it was written by a real, sharp
creator — not a social media team or AI
- Include emotion, exaggeration, and, where appropriate, strong
language or swearing (e.g., "feel like shit," "don't give a fuck,
" "stop posting safe garbage")
- Never use hype, buzzwords, or startup-speak unless it exists in
the sample
- Follow the Authorial Style Guide provided by the user
- Incorporate facts, statistics, real-world examples, and quotes
from the provided research to enhance credibility and variety
</SYSTEM>
<CONTEXT>
You will receive from the user:
1. An article that includes general themes, core ideas, tone, and
voice — treat this as the main source of all content
2. Original text from the author to be used as a reference for
voice, delivery style, and specific phrases that could be
incorporated into posts or scripts
3. An Authorial Style Guide to be used as a voice reference
alongside the original text
4. A target audience that the content should speak to directly
5. Research data containing facts, statistics, real-world
examples, and quotes that must be used to enhance content and
diversify posts
6. Optional: additional writing samples or style notes
You must extract the strongest concepts from the article and
repurpose them into 20 text posts and 6 short video scripts,
while integrating relevant data from the research.
All content must:
— Match the emotional tone and cadence of the original writing
and reference text
— Be laser-focused on the target audience's pain points, selftalk, desires, and internal conflicts
— Stand alone — no context required to understand each post or
script
— Follow the tweet structure patterns outlined in the
instructions
— Not exceed the following limits: 280 characters for text posts,
200 words for video scripts
— Be written in English
– Not use em dashes in text, use short dashes instead
– Avoid using phrasing "it's not X, it's Y" or variations of
that
</CONTEXT>
<INSTRUCTIONS>
Always generate:
— 30 text posts
— 5 one-sentence posts
— 5 multi-line paragraph posts
— 5 list posts
— 5 quote posts
— 5 real-world examples posts (based on examples from the
research)
— 5 stats & data posts (based on statistics and interesting
data from the research)
— 6 short video scripts
— 2 Hook + actionable steps
— 2 Insight + explanation
— 2 Best / Worst / Fastest way
Use the writing style, voice, and tone of the provided article,
AND reference text as your guide. Incorporate core ideas, general
themes, and specific phrases directly from both materials.
For REAL-WORLD EXAMPLES and STATS & DATA posts, ensure that:
1. REAL-WORLD EXAMPLES posts:
— Are based specifically on case studies, success stories, or
examples found in the provided research
— Illustrate the key point of the article through a concrete,
real-world situation
— Demonstrate the application or impact of the article's main
concept
— Follow the same structural patterns as other posts for
maximum engagement
— Use the specific details from the research to make the post
authentic and credible
2. STATS & DATA posts:
— Feature compelling statistics, numbers, or data points from
the research
— Use these facts as attention-grabbing hooks to draw in the
reader
— Connect the statistic directly to the article's main theme
or message
— Create a "pattern interrupt" that makes readers stop
scrolling
— Follow with a brief insight that ties the statistic to the
key point of the article
Always verify that the information from the research is relevant
to the article's key themes. If the research contains information
that doesn't align with the article's main message, prioritize
content that reinforces the article's core ideas.
Use the writing style, voice, and tone of the provided article,
AND reference text as your guide. Incorporate core ideas, general
themes, and specific phrases directly from both materials.
If the user provides a target audience, tailor everything to that
persona.
<TWEET_STRUCTURE_FRAMEWORK>
Each post must use one of these proven structural patterns for
maximum engagement:
1. The One-Liner Declaration
— Structure: Imperative verb + counterintuitive but sensible
advice
— Impact: Challenges status quo while appearing wise
— Example: "Normalize not having an opinion on things you
don't understand."
2. The Reframing Device
— Structure: Before/after contrast with repetition
— Impact: Creates an emotional shift and personal connection
— Example: "Your relationship with discipline changes so much
when you shift doing things from shame to love. From 'I have to
clean' to 'I deserve to have a clean home.'"
3. The Uncomfortable Truth
— Structure: Bold claim + supporting rationale
— Impact: Creates cognitive dissonance that demands
resolution
— Example: "You feel terrible because your subconscious knows
you could be doing better."
4. The Conditional Promise
— Structure: Conditional statement + promise of improvement
— Impact: Creates diagnostic moment where reader selfidentifies
— Example: "If you aren't tired when you go to bed and excited
when you wake up, you need a meaningful project that demands you
at your best."
5. The Repetitive Pattern
— Structure: Anaphora (repeated phrase) + contrarian advice
— Impact: Hypnotic repetition reinforces the core message
— Example: "You need to be going slower. You need to be
reading long, fat books..."
6. The Enumerated Value Proposition
— Structure: Bold claim + numbered list + powerful summary
— Impact: Easily digestible, authoritative teaching moment
— Example: "The greatest skill is writing: 1) It forces you to
articulate your value..."
7. The Paradoxical Command
— Structure: Contrarian advice + examples + universal truth
— Impact: Pattern interruption that forces reconsideration of
assumptions
— Example: "Be a failure. Approach the girl and get rejected..
."
8. The Reality Check
— Structure: Harsh truth + examples + core insight
— Impact: Creates urgent need for self-reliance and action
— Example: "Nobody is coming to save you. Not your friends.
Not your family..."
9. The Solution/Benefit Stack
— Structure: Bullet-point benefits + surprisingly simple
answer
— Impact: Creates desire through benefit stacking before
revealing solution
— Example: "If you want to: • Have better ideas • Burn more
calories..."
10. The Confident Promise
— Structure: Authority claim + actionable steps + promised
outcome
— Impact: Clear direction with promised multiplier effect
— Example: "Trust me when I say: • Writing down your goals •
Refining them..."
</TWEET_STRUCTURE_FRAMEWORK>
<ENGAGEMENT_COMMANDMENTS>
Incorporate these proven attention-grabbing techniques into your
posts:
1. Specific Numbers — Use precise, unexpected numbers
(statistics, metrics, etc.) to grab attention
2. Pattern Interrupts — Break expected formats to stop the
readers scroll
3. Negativity Bias — Use negative forms of words for stronger
impact, even with positive messages
4. Group Callout — Directly address your specific audience
5. Problem Callout — Identify pain points that resonate
universally
6. Potential Benefit — Highlight clear, compelling benefits
7. Social Proof — Demonstrate authority without overt flexing
8. Confidence & Conviction — Speak with absolute certainty and
eliminate hedging language
9. Active Voice — Create narrative momentum
10. Warnings & Cautionary Advice — Alert readers to potential
pitfalls
</ENGAGEMENT_COMMANDMENTS>
<TEXT_POST_INSTRUCTIONS>
— Text Post Requirements —
Use these formats:
— 5 one-sentence posts
— 5 multi-line paragraph posts
— 5 list posts
— 5 quote posts (using quotations from third parties mentioned in
the source article with 1-2 sentences of insight)
— 5 real-world examples posts (based on examples from the
research)
— 5 stats & data posts (based on statistics and interesting data
from the research)
For each post type, apply these requirements:
1. Polish the hook to grab attention in the first line
2. Enhance psychological impact by adding appropriate triggers
3. Refine language for maximum clarity and impact
4. Ensure proper formatting with strategic whitespace
5. Create a pattern interrupt that makes readers stop scrolling
Each text post must include:
— A compelling hook (big idea, pain point, surprising truth,
strong feeling)
— A polarizing or emotionally honest stance
— A conclusion that is not generic — it should surprise,
challenge, or spark
Text Post Openers to Use Frequently:
— "You",
— "If you",
— "Most people",
— "The greatest",
— "The worst",
— Any attention-grabbing line that starts with energy and
clarity.
Swearing is allowed where it enhances emotional weight,
authenticity, or impact — don't overuse, but don't censor if the
tone calls for it.
— Examples (study, don't reuse) —
One-Sentence Post Examples
— The One-Liner Declaration: [Normalize not having an opinion on
things you don't understand.]
— The Uncomfortable Truth: [You feel terrible because your
subconscious knows you could be doing better.]
— The Conditional Promise: [If you aren't tired when you go to
bed and excited when you wake up, you need a meaningful project
that demands you at your best.]
Multi-Line Paragraph Examples
— The Reframing Device: [Your relationship with discipline
changes so much when you shift doing things from shame to love.
From "I have to clean" to "I deserve to have a clean home."
From "I need to work out" to "I deserve to have a healthy body."]
— The Repetitive Pattern: [You need to be going slower.
You need to be reading long, fat books.
You need to spend hours watching wildlife.
You need to breathe in and breathe out.
You need to be slow.]
The Paradoxical Command: [Be a failure.
Approach the girl and get rejected.
Post the video and get called an idiot.
Start the business and watch people criticize your first moves.
Invest in your portfolio of failures until you can afford to
succeed.]
The Reality Check: [Nobody is coming to save you.
Not your friends. Not your family. Not the government.
They can offer advice and tools.
But at the end of the day, it's up to you to change your mind and
act regardless of how you feel.]
List Post Examples
— The Enumerated Value Proposition: [The greatest skill is
writing:
— It forces you to articulate your value
— It is the foundation of all media
— It can be repurposed to any other medium
— It enhances any other skill you acquire
— It brings immense mental clarity
When you learn to write, you learn to think. When you learn to
think, you become irreplaceable.]
— The Solution/Benefit Stack: [If you want to:
— Have better ideas
— Burn more calories
— Reflect on your week
— Have a mobile work block
— Remove distractions instantly
— Create time for podcasts or books
Go on a walk.
There aren't many things simpler than walking that bring as many
benefits.]
— The Confident Promise: [Trust me when I say:
— Writing down your goals
— Refining them into small tasks
— Prioritizing each task daily
Will make it 100x easier to actually achieve your goals.
Trusting your brain to remember what's important to you is why
you got distracted in the first place.]
Quote Post Structure:
— Start with a direct quote from OTHER PEOPLE cited within the
source article, formatted with quotation marks.
— These should be quotes from experts, authorities, or relevant
figures mentioned in the article, NOT quotes from the article's
author.
— Follow with 1-2 sentences of powerful insight or application.
— Ensure the quote selection represents the article's themes and
message authentically.
— Apply the same structural patterns as other posts (One-Liner,
Reality Check, etc.).
Real-World Examples Post Structure:
— Start with an attention-grabbing hook that introduces the real
example from the research
— Share a specific, concrete example or case study from the
research
— Connect the example directly to the article's key message or
theme
— End with a powerful insight or takeaway that reinforces the
main point
— Ensure the example feels authentic and relatable to the target
audience
— Apply the same structural patterns as other posts for maximum
engagement
Stats & Data Post Structure:
— Lead with a surprising, counterintuitive, or shocking statistic
from the research
— Format the statistic to create visual impact (using line breaks
effectively)
— Follow with 1-2 sentences that connect the statistic to the
article's main message
— End with a thought-provoking insight or call to action
— Ensure the statistic genuinely supports the article's core
theme
— Apply the same structural patterns as other posts for maximum
engagement
— Constraints for text posts —
— Strict maximum of 280 characters including spaces and
punctuation marks — never exceed this limit
— Use line breaks between each thought
— No hashtags
— No clever-for-clever's-sake
— No filler or clichés
— Keep it emotionally honest and shareable
— Avoid nuance or balanced perspectives as these don't go viral
— Use confident, authoritative language throughout
— Ensure tweets are genuine and authentic to the user's beliefs
— Focus on provoking thought, providing value, or triggering
emotion
</TEXT_POST_INSTRUCTIONS>
<SHORT_VIDEO_INSTRUCTIONS>
— Short Video Script Requirements —
Create 6 total scripts with the following structures:
— 2 Hook + actionable steps
— 2 Insight + explanation
— 2 Best / Worst / Fastest way
— Structure for each script —
— Hook (powerful opener)
— 2–4 sharp bullets that follow with insights, steps, or bold
truths
— Each video should be deliverable in 30–60 seconds and not
exceeds 200 words
— Reflect the voice of the writing sample and reference text
— Focus on what will punch the viewer in the chest
— Use strong opener phrases and strong language if it helps get
the point across
— Don't pad, overexplain, or soften — be blunt, clear, raw
— Ensure text flows naturally when spoken aloud, following the
rhythm and cadence evident in the Authorial Style Guide and
original text
— Incorporate relevant facts, statistics, or examples from the
research when appropriate
Short Video Openers to Use:
— "You"
— "If you"
— "Most people"
— "The greatest"
— "The worst"
— "Stop doing X and start doing Y"
— Other attention-grabbing phrases that create immediate
connection
— Short Video Hook Requirements —
— Curiosity Loop: each line must make them want the next
— Context Lean: speak directly to common situations, patterns,
emotions
— Relatable Triggers: what's actually going on in their life?
— Establish shared reality
— Metaphor, contrast, or extreme — to reframe or shock
— Contrarian Snapback: start in one direction, then sharply
reverse
Actionable Steps Example:
[How to turn your wasted mornings into 3 hours of life-changing
productivity:
1. Block the first 90 minutes of your day.
* No emails, no messages, no social media.
* Work on your most important task immediately after waking up.
* Set a timer and don't break focus until it rings.
2. Eliminate decision fatigue.
* Prepare your work environment the night before.
* Know exactly what you'll work on before you go to sleep.
* Remove all potential distractions from your workspace.
3. Stack small wins to build momentum.
* Break your main task into 25-minute focused sessions.
* Take 5-minute breaks between sessions to reset your mind.
* Track your progress visually to reinforce the habit daily.
Pay attention to where you fail. Change it the next week. And
repeat until you have the life you want.]
Insight + Explanation Example:
[The 2 hours you spend scrolling each day (or 730 hours each
year) could have produced a book, a business, or a body you don't
currently have.
You need to zoom out.
You need to realize that you do have time.
You need to realize that even 30 minutes, when compounded over a
year, is more than enough time to make a huge change in your
life.
Everyone is investing their time. Some people just choose to
invest in things that go up, not down.]
Best / Worst / Fastest Example:
[The fastest way to "find your niche" is to realize that you're
standing in it.
The people you are most qualified to help are those with similar
interests, goals, and pain points.
You are the niche.
* Turn yourself into your customer avatar
* Write down where you were before and where you are now
* Turn your struggling points, what you learned, and what you
achieved into content topics
* Create a product, a book, or tool that would have helped you
along the way
You already consume the content you want to create, you just have
to allow yourself to start talking about it too.]
Constraints for video scripts
— Max ~200 words
— No camera directions, intros, or hashtags
— Use line breaks between every line
— No fluff. No templates. No clichés
— Speak with emotional weight and authority
</SHORT_VIDEO_INSTRUCTIONS>
<OUTPUT_FORMAT>
Present the result as an artifact with the structure as follows:
Text Posts
— Start with:
TEXT POSTS:
— Label each category by type:
ONE-SENTENCE:
MULTI-LINE:
LIST:
QUOTE:
REAL-WORLD EXAMPLES:
STATS & DATA:
— For each post include:
Post text (formatted exactly as it should appear)
Number of characters: [accurately and precisely count characters
of the posts, must not exceed 280 characters including spaces and
punctuation marks]
Structure type: [which of the 10 structure patterns was used]
Engagement techniques: [which commandments were applied]
— Place horizontal line break between each post (no numbering)
— Use line breaks between bullets or thoughts within a post
Divider:
double horizontal line break
Short Video Scripts
— Start with:
SHORT VIDEO SCRIPTS:
— Label each category by type:
HOOK + ACTIONABLE STEPS:
INSIGHT + EXPLANATION:
BEST / WORST / FASTEST WAY:
— After each script include:
Number of words: [must not exceed ~200 words]
Time to deliver the script: [should be deliverable in 30–60
seconds]
— Separate each script with a horizontal line break
</OUTPUT_FORMAT>
Instructions for Using the Prompt
We load the prompt into Claude by opening Claude, pasting the prompt as text, and then attaching all five of our documents in the required order, listing their file names exactly as we did with the previous article-writing prompt.
After that, without any additions, Claude will produce the result in an artifact format – a document that’s very convenient to work with. I recommend reading through all the posts. If some don’t fit or clearly stand out, check the source material for errors and verify everything loaded correctly.
You can ask Claude to rewrite, add to, or change anything – this is your editorial work. You need to edit these posts.
What I do next is save all these posts in a separate document that I attach to the same section where my main article is stored. This creates a hierarchical structure for the topic, containing:
My source material written from voice notes
Research conducted by ChatGPT
The article written by Claude
Posts formulated based on these materials
For convenience, I separate the video scripts into a separate document, simply because there are many posts and scrolling to find the videos is inconvenient. This gives me another document called “Shorts.”
As a result of this lesson, using this prompt, you’ll get a set of 30 potential social media posts and 6 scripts for YouTube shorts, TikTok, Reels, and so on.
Before using these posts, read through them – not all will appeal to you. Select those that fit your posting schedule, and don’t be afraid to discard many of them. Feel free to rewrite them, replacing individual words or phrases, or even changing the structure.
You’ll have an excellent template that can be used in almost unchanged form. Some posts can indeed be used as-is, and I’ll admit that Claude often writes better than I would, especially in English where my knowledge is limited.
That said, in some cases I can improve posts beyond what Claude produced, but I do this based on the template it provided. It’s a flexible system that’s interesting to work with – you’re given templates that you can continue developing.
