Skills

Skills you can steal

Markdown skill files that teach an AI assistant how to do one job properly — the ones actually used to run this business. Each drops straight into Claude, ChatGPT, Hermes, or whatever you run locally. Free, editable, no signup.

GTM · Paid skills

SEO / GEO Optimization

A separate set of skills for ranking in Google and getting cited by ChatGPT, Perplexity and AI Overviews — keyword research, topic scoring, GEO optimization and content auditing. These ones are for sale rather than free.

  • SEO
  • GEO
  • AI visibility
  • Paid
Browse the store

Ideation · Creative

ADHD

Parallel divergent ideation: fan five isolated agents out under different cognitive frames, score and cluster what comes back, prune the traps, then deepen the best three into buildable sketches with an actual opinion.

  • Brainstorm
  • Multi-agent
  • Architecture
  • Anti-textbook
Open skillDownload .md

GTM · Data

Clay API

Clay's whole programmatic surface in one reference: which of the three surfaces (REST, MCP, CLI) can actually do the thing, the non-standard auth header, what is UI-only — and the two rules that stop you spending someone's credits by accident.

  • Clay
  • Enrichment
  • REST
  • MCP

GTM · Strategy

GTM Expert

Six practitioner corpora — Crawford, Lieben, Voje, Hérubel, Stam, Vacca — routed by what the question is actually about. Returns their real steps and real numbers, flags where they contradict each other, and labels every claim measured, asserted or marketing.

  • GTM
  • ICP
  • Outbound
  • Positioning

Research · GTM

ICP Research

Exa finds the page that actually states the ICP, Firecrawl reads it properly, then the client logo wall gets resolved to real headcounts — so company size is observed, not guessed from a price. Every field carries a verbatim quote or stays empty.

  • ICP
  • Exa
  • Firecrawl
  • Research

AI · Cost

AI Cost Optimizer

Before you run a model over 10,000 rows, price the job. Reads OpenRouter's live rankings and live pricing, costs your actual run in dollars, and puts the open-weight labs next to the frontier ones — the comparison a from-memory shortlist always skips.

  • OpenRouter
  • Cost
  • Models
  • Bulk jobs

Research · Competitive

Ad Library Recon

Who a company targets, not who it has already won. Reads live ads out of the LinkedIn, Meta and Google ad libraries with a real browser — they are JavaScript apps that block scrapers outright, so this is the only way in.

  • Ads
  • ICP
  • Competitive
  • Browser
Open skillDownload .md

Data · API

Google Sheets API

All 17 core and values methods, all 74 batchUpdate request types, and the OAuth flow from a client ID to a working refresh token — plus the three behaviours that quietly cost people data: append writes somewhere else, reads come back ragged, and the wrong field mask erases cells.

  • Sheets
  • Google API
  • OAuth
  • Reference

Data · API

Supabase API

PostgREST, direct Postgres for bulk loads and DDL, and the four key types — including which one fails silently. Written around the three things that make inserts look successful when nothing landed: the 1000-row cap, RLS returning empty arrays instead of errors, and DDL that REST cannot do at all.

  • Supabase
  • PostgREST
  • Postgres
  • Reference

Strategy · Build

Project Approach Advisor

Before you commit to a stack: researches how comparable projects were actually built, prices the operating cost of each option, and comes back with one recommended approach rather than a menu of them. For when there is no look-alike project to copy.

  • Architecture
  • Research
  • Stack
  • Cost

Ops · QA

Human Review Packet

The difference between a review that happens and one deferred session after session: every machine-made claim shipped next to its source URL and a local snapshot of the text it came from, so the reviewer checks the work instead of redoing it.

  • Review
  • QA
  • Data
  • Verification

Research · Papers

NeuroArxiv

Checks arXiv before you invent an algorithm. Reads the prior art in parallel, scores and clusters it, then converges on one path with citations, a first step, and the pitfalls whoever published it already walked into.

  • arXiv
  • Architecture
  • Prior art
  • Multi-agent
Open skillDownload .md

Ops · Data

Versioned Artifacts

Never overwrite a file a human has already been given. The revision gets a new dated filename and the old one stays put, so the person reading yesterday's copy and the agent reading today's are never quietly looking at different data under one name.

  • Data
  • Versioning
  • Handoff
  • QA

Ops · Discipline

Unlazy

Anti-laziness execution discipline for agents that report done at 80 percent. v2 stopped trusting promises: acceptance gates go in a file before work starts, each one carries a runnable check, and completion is proven against the ledger instead of asserted.

  • Discipline
  • Gates
  • Autonomous runs
  • QA

Design · Frontend

Design Taste

Anti-slop frontend for landing pages, portfolios and redesigns. Reads the brief before touching a dial — page kind, vibe words, references — then pulls only the rules that fit, so the result stops looking like the template every model defaults to.

  • Design
  • Frontend
  • Landing pages
  • Anti-slop
Open skillDownload .md