A prompt is a one-time thing.
You write it. Claude uses it. The session ends. Next time you open Claude, it is gone. You write it again. Slightly different. Slightly worse. You spend the first five minutes of every session reconstructing context you already built last week.
A skill runs forever.
You build it once. Claude loads it automatically. Every session starts with Claude already knowing how to do the specific thing you need — in your voice, at your standard, with your rules baked in.
Most Claude users have never built one.
Save this. You will read it twice.
What a Skill Actually Is
A skill is a .md file that gives Claude a permanent job description for a specific task.
Not a prompt. A job description. The difference: a prompt tells Claude what to do once. A skill tells Claude who to be every time a specific task appears.
You save it in Claude once. After that, Claude loads it automatically when the task matches. You stop rebuilding context. You stop getting generic output. You stop starting from zero.
The 20 skills below cover every category a serious builder works in. Copy the ones you need. Save them. Start every session with Claude already knowing how to work.
How to Install Any Skill in This List
Claude.ai: Settings → Sub-Agents → Add. Paste the skill block. Name it. Save.
Claude Code: Create .claude/agents/ in your project root. Save each skill as skillname.md. Claude finds them automatically.
Claude Desktop: Settings → Sub-Agents → Add. Same as Claude.ai.
Build once. Run forever.
Category 1: Content & Writing
01. The Hook Forge
What it does: Takes any topic, angle, or draft opening and generates 10 hooks across different psychological triggers — curiosity, loss, contrast, specificity, controversy. You pick the one that performs. You stop staring at a blank first line.
Fire it when: You have an idea and no opening line. Run this before you write a single word of the piece.
02. The Voice Locker
What it does: Analyzes your writing samples and locks your voice permanently into Claude. Every piece Claude writes after this sounds like you — not like a polished AI assistant, not like a corporate blog, like you.
Fire it when: The first time you set up a new Claude Project. Run this before anything else. Every piece of content that comes after will be better for it.
03. The Thread Architect
What it does: Takes any idea, article, or argument and builds it into a Twitter/X thread with the structural precision that actually performs — hook, tension, payoff, CTA. Not a summary of the idea. A thread built to be read to the end.
Fire it when: You have a piece of content and you want a thread version that performs — not just a list of points from the article.
04. The Repurpose Engine
What it does: Takes one long-form piece and produces every short-form version at once. Thread, LinkedIn post, newsletter section, three standalone hooks, one quote graphic caption. One input. Six outputs. No reformatting by hand.
Fire it when: You finish any long-form piece and need distribution without spending three hours reformatting.
05. The Headline Lab
What it does: Takes any article or content idea and generates 15 headlines across different proven formats — how-to, number lists, question, contrast, secret, warning. Scored and ranked. You pick from the top three and publish with confidence.
Fire it when: You have finished a piece and you are guessing at the title. Headlines are not decoration. They are the difference between 200 readers and 20,000.
Category 2: Research & Analysis
06. The Brief Builder
What it does: Takes any topic and builds a research brief that a junior analyst could execute. Sources to check, questions to answer, signals to look for, and the one finding that would change everything. Turns a vague research request into a structured investigation.
Fire it when: You are about to research something and you have a topic but no structure. Brief first. Research second. Always.
07. The Contradiction Finder
What it does: Takes any body of information, set of sources, or argument and finds where the sources contradict each other. Surfaces the real complexity underneath the consensus. Most research flattens contradictions. This one finds them.
Fire it when: You are building an argument from research and you want to know where your sources are weakest before someone else finds it.
08. The Signal Scanner
What it does: Takes a week's worth of industry news, posts, or content and separates signal from noise. What actually matters versus what got attention. Patterns emerging versus one-time events. The brief you wish someone sent you every Monday.
Fire it when: You have a backlog of industry content and you need to know what actually matters before your week starts.
09. The Assumption Auditor
What it does: Takes any plan, argument, or decision and surfaces every assumption baked into it — including the ones so obvious nobody thought to question them. The assumptions you do not know you are making are the ones that kill projects.
Fire it when: You are about to launch something, pitch something, or commit resources to something. Run this the day before. Not after.
10. The Source Ranker
What it does: Takes a list of sources on any topic and ranks them by credibility, recency, and relevance. Flags which ones to trust, which to verify, and which to cut entirely. Stops you from building arguments on weak foundations.
Fire it when: You have gathered sources and you are about to start writing. Rank them first. Build on the right three. Cut the rest.
Category 3: Business & Operations
11. The SOPWriter
What it does: Takes any process you describe — in rough notes, bullet points, or a voice memo transcript — and turns it into a clean, executable SOP. With owners, tools, steps, and the one step that must never be automated.
Fire it when: A process exists in your head or in messy notes and needs to become something someone else can execute.
12. The Decision Framer
What it does: Takes any decision you are stuck on and reframes it into a structured choice with clear criteria, real options, and the information you actually need before deciding. Stops you from making decisions by feel on things that deserve a framework.
Fire it when: You have been thinking about a decision for more than a week without deciding. That is the sign you do not have a framework. Run this first.
13. The Meeting Extractor
What it does: Takes any meeting transcript or rough notes and returns the only four things that matter: what was decided, who owns what, what is blocked, and the one thing nobody said out loud that will affect everything.
Fire it when: Every meeting ends. No exceptions.
14. The Pricing Stress Tester
What it does: Takes your pricing model and stress tests it from three directions — the customer who thinks it is too expensive, the customer who thinks cheap means bad, and the competitor who is about to undercut you. Returns what your pricing is actually communicating and what to change.
Fire it when: You are about to launch new pricing or you have been losing deals and you are not sure why.
15. The Offer Sharpener
What it does: Takes any offer — a product, service, or proposal — and sharpens it until the value is impossible to misunderstand. Finds the gap between what you think you are selling and what the buyer thinks they are buying.
Fire it when: You are describing your product or service and people keep saying "interesting" and not buying. That is a gap problem. This finds it.
Category 4: Coding & Dev
16. The Code Explainer
What it does: Takes any piece of code — regardless of language or complexity — and explains it in plain language at two levels: what it does for someone who needs to use it, and why it is built this way for someone who needs to change it.
Fire it when: You inherit code you did not write, onboard a new developer, or prepare to modify something you have not touched in six months.
17. The PR Reviewer
What it does: Reviews any pull request diff and returns the five things that matter: bugs, missing tests, security issues, style violations, and the one architectural decision that deserves a real conversation before merging.
Fire it when: Any pull request is ready for review. Run this first. It catches 80% of issues in 90 seconds.
18. The Debug Partner
What it does: Takes any error, stacktrace, or unexpected behavior and walks through the diagnosis systematically. Root cause first. Band-aid fixes flagged and rejected. Regression test included.
Fire it when: You have been staring at a bug for more than 20 minutes. Stop guessing. Run this.
Category 5: Strategy & Thinking
19. The Second Order Thinker
What it does: Takes any decision, action, or trend and maps the second and third order consequences — the effects of the effects. Most people think one move ahead. This skill forces three.
Fire it when: You are about to make a decision that affects more than the next 30 days. Or when you are analyzing a trend everyone else is treating as a first-order story.
20. The Mental Model Applier
What it does: Takes any problem or decision and applies the three most relevant mental models to it. Not a list of mental models. The right ones for this specific situation, applied to produce a specific insight.
Fire it when: You have been thinking about a problem from the same angle for too long. Run this when you need a perspective shift, not more information.
The Uncomfortable Truth
You have been building prompts that expire.
Every session you start from zero. Every generic output is the result of Claude not knowing who you are, how you think, or what standard you hold. You have been fixing that gap manually, every time, with a few lines of context that disappear when the chat ends.
Skills fix it permanently.
Twenty skills. One afternoon of setup. Every session after that starts with Claude already knowing how to work with you — at your standard, in your voice, with your rules already loaded.
The builders producing ten times the output are not smarter. They built the skills once and stopped rebuilding them every session.
Open the folder. Paste the files. Stop starting from zero.
Follow @sairahul1 for more skills, sub-agents, and systems that make Claude work the way you actually think.
TL;DR
20 skills. Five categories. Build once. Run forever.
Content & Writing
01. Hook Forge — 10 hooks across 10 psychological triggers before you write a word.
02. Voice Locker — locks your voice permanently so Claude stops writing like an AI.
03. Thread Architect — builds threads for retention, not just coverage.
04. Repurpose Engine — one input, six platform-native outputs.
05. Headline Lab — 15 headlines, scored and ranked, top 3 to test.
Research & Analysis
06. Brief Builder — turns vague research requests into structured investigations.
07. Contradiction Finder — surfaces where your sources disagree before someone else does.
08. Signal Scanner — separates what actually matters from what just got attention.
09. Assumption Auditor — finds the invisible assumptions before they kill the plan.
10. Source Ranker — ranks every source before you build an argument on it.
Business & Operations
11. SOP Writer — rough notes in, executable process out.
12. Decision Framer — reframes stuck decisions into something you can actually choose.
13. Meeting Extractor — decisions, owners, blockers, and the unspoken issue. Nothing else.
14. Pricing Stress Tester — stress tests pricing from three buyer perspectives.
15. Offer Sharpener — finds the gap between what you think you are selling and what buyers hear.
Coding & Dev
16. Code Explainer — two-level explanation for users and modifiers.
17. PR Reviewer — bugs, tests, security, style, and the one decision worth a conversation.
18. Debug Partner — root cause diagnosis. No band-aids.
Strategy & Thinking
19. Second Order Thinker — maps consequences three moves ahead.
20. Mental Model Applier — the right frameworks for this specific problem, applied specifically.
The prompt expires. The skill runs forever.
