You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
35 Claude skills distilled from EP's (@eptwts) Signal Knowledgebase — "all the signal from my 14,000 tweets" — at eptwts.com, plus his partnership-program page and a sample of his recent/top X posts.
EP's thesis, and this repo's organizing idea: all roads lead to distribution. A 10x worse product with the right distribution beats the best product with none. The skills cover the whole loop he describes — picking a market, designing an offer, driving traffic, converting in owned channels, using AI as a multiplier, and executing.
Everything here is EP's opinionated operator heuristics, attributed to him, not universal fact. Tactics he flags as dated carry a "practice might be outdated" note. ToS-violating or manipulative tactics he describes are kept out of every skill's workflow and appear only in the verbatim references, marked "don't run."
Layout
skills/<skill-name>/
├── SKILL.md # frontmatter + principles, workflow, templates, anti-patterns
└── references/source-lessons.md # the verbatim source lessons the skill is built from
sources/ # full scrape of eptwts.com (8 chapters), partners page, X sample
SKILL_SPEC.md # the authoring spec every skill follows
Using the skills
Claude Code / Cowork: copy any skills/<name>/ folder into your skills directory (e.g. ~/.claude/skills/ or a project's .claude/skills/). Each SKILL.md is self-contained; the references/ file is loaded only when needed.
Claude.ai: zip a skill folder and upload it under Settings → Skills, or run python -m scripts.package_skill skills/<name> from the Anthropic skill-creator to produce a .skill file.
Skills reference each other in their "Read next" sections — e.g. distribution-first-strategy hands off to offer-design-and-buyer-psychology, which hands off to funnels-and-owned-channels.
Pick and walk an end-to-end online business playbook matched to the user's constraints (skills, capital, willingness to show their face, coding ability).
Decide how much to build on a personal brand vs. faceless or operator-behind-creators models, what to reveal publicly (niche, earnings, methods, client results), how to vet gurus and communities, and how to convert volatile online income into durable wealth.
Audit and optimize an X (Twitter) post or account against the ranking signals the open-source X algorithm actually weights (retweets/quotes, follows, dwell time, profile clicks, replies, negative signals, tweepcred, author-diversity decay).
Build a phased 0-to-10k follower plan for a new or tiny X (Twitter) account using EP's playbook — big-account engagement as the distribution unlock, the affiliate route, experience-based value posts, lead magnets under 10k, follow lists, intro tweets, profile/bio rules, and weekly cadence.
Define who a creator's content is for, set content pillars that attract buyers rather than spectators, and run a pre-publish checklist (provoke / spark / start, reader-as-protagonist, no self-documentation early, honesty over flex).
Choose the platform mix for a specific offer (X vs YouTube vs TikTok/Reels/Shorts) using EP's traffic-quality map, and run a YouTube channel launch and diagnosis procedure (first 6-10 video test, search vs browse, never linking from outside, permanent-VSL structure, Shorts related-video trick).
Run platform-specific growth funnels the whitehat way — Instagram comment-to-DM funnel, 7-second loop reels, localized town/community pages run by real hires, clipping partnerships with permission, spectacle/trend-adjacent content — and explain why EP's grey-hat versions (bought followers, account farms, fake persona pages, proxies) are not worth running.
Turn a raw idea into a post, thread, or piece of content that gets reach - hook it, dumb it down, structure it for scannability, and pick the right format/platform.
Write persuasive, conversion-focused copy (landing pages, VSL scripts, sales emails, ad copy, lead magnets, product-promoting content) by mapping the customer's desire chain, mining real customer language from Reddit/Quora/reviews, and structuring content like a VSL.
Build a reusable brand-voice profile from a corpus of someone's writing and run a daily writing loop where a human braindumps, AI formats, and the human rewrites in their own voice - so posts never read as AI-generated.
Design a monetization funnel that routes social/content traffic into an owned channel (email list or Telegram) before a sales page, and decide how to pair a traffic source with an offer.
Design and price an info-product ladder (low-ticket front end → paid community → high-ticket done-with-you/service) that sells a transformation rather than information.
Launch and run a paid community - platform choice (Telegram vs Discord vs Whop), pricing gate, structure, onboarding, automated upsells, support, and retention metrics - plus the payment and platform mechanics around it.
Pick a niche by viewer value and run one of EP's complete step-by-step online business playbooks (minimal online business, $0-to-$10k insecurity play, level-0 baseline, faceless niche content, influencer piggyback, faceless trading brand, operator revshare, Telegram crypto channel, college pages, UGC retainer, beginner affiliate, local events, CPA training ground).
Evaluate whether an info offer, course, mentorship, or "guru" is legitimate - as a buyer deciding whether to pay, or as a seller deciding how to market and whom to accept.
Plan and execute a first online income path via clipping/UGC and affiliate marketing - learn traffic as the transferable skill before building any product.
Run EP's concrete affiliate-traffic plays as step-by-step checklists - clipping-and-affiliate at scale, the beginner content-boost-for-promotion deal, broke-start replication with an owned channel, the YouTube-to-X transcript engine, language-arbitrage promo videos, AI slideshows, and AI angle-mining - plus a "dead/patched, don't run" register.
Build, store, and stack reusable JSON context profiles (business, ICP, brand voice, marketing strategy, personal context) so every AI session starts with your situation instead of from zero.
Turn a vague or messy request into a precise, high-yield prompt using role assignment, context collection, step-by-step instructions, example output, rules, multi-role debate, recursive prompting, and session serialization.
A library of ready-to-paste anti-sycophancy advisor prompts (brutally honest strategic advisor, first-principles problem solver, accountability manipulator, life optimizer, LogicCore, goal-to-checklist, journal profiling, and more) and a picker for which one fits the user's situation.
Ready-to-paste interview prompts for each stage of a business, from scoring an idea backlog and matching a business model, through positioning, ICP and marketing-strategy JSON, Reddit pain-point and distribution-channel deep research, to an execution checklist.
Learn any book, video, PDF, course, or skill faster with AI by extracting only what applies to your situation, running interactive tutoring, and building personalized curricula.
Extract a reusable JSON visual style profile from reference images and use it to generate on-brand thumbnails, ads, infographics, UI, and decks with image models.
Produce AI-generated UGC and short-form video ads by researching winning videos, analyzing them with Gemini, rewriting the context for your product, and generating with image-to-video models, then editing.
Set up a coding-agent harness and a continuously-fed second brain (Obsidian + Claude Code, drop-folder pipelines, RAG over conversation data, always-on server agent) for an operator, and pick the right model and tool per task.
Runs a self-audit, builds a skill-stack plan, and sets a 90-day skill sprint for someone deciding what to do with their career or online-business ambitions.
Evaluates whether to trust a piece of advice, a guru, a course, or an offer, and sanity-checks personal finances for people with volatile internet income.
Sources
File
Content
sources/00-eps-x-signal-official.md
EP's own "Download everything as Markdown" export from eptwts.com (693 lessons) — the canonical source