Makerskills: AI Agent Skills for Solo Operators, Not Just Coders

Most AI agent skills built for Claude Code are about writing and shipping code. Makerskills is a rare entry aimed at the other half of running a one-person business: the decisions, the research, the content calendar, the cash flow. It's a collection of AI agent skills for solo operators who use an agent as a working partner for the craft of the job, not just the codebase.

Built by Corey Haines, founder of Conversion Factory and Magister, the repo has crossed 561 stars and 46 forks since it went public in early June 2026, with commits landing as recently as August 1. It's MIT licensed, free, and open source.

What's actually in the repo

Makerskills ships 19 skills, grouped into four families. Each one is a SKILL.md file plus supporting logic, installable as a Claude Code plugin (it also works with Codex and Cursor):

Meta skills, for extending the system itself

• skillify creates, adapts, or updates skills, including porting an external skill in with a license check and attribution
• toolify wires up an API, MCP server, or third-party integration into a Next.js or Rails project
• loopify sets up a cron job or recurring agent loop with idempotency and bail-out conditions built in

Decision and strategy skills

• decide runs the 37signals decision framework, a set of questions triaged down to 6 to 8 per decision, and archives the outcome with a date to revisit it
• business-brainstorm pressure-tests a new business or product idea across nine dimensions
• domain runs an 11-step domain hunt combining whois, Domainr, Namecheap, and a trademark screening pass
• deep-research does multi-source research with citations and an archive
• unstuck classifies what kind of wall you've hit (a bad assumption, a framing problem, a gatekeeper, a missing tool) and runs lateral-thinking techniques against it

Knowledge and content-consumption skills

• second-brain is a capture-compile-query-lint workflow over a personal markdown vault
• company-brain is the team-scope version, with source attribution and a review pass so stale information doesn't poison later answers
• read-book produces chapter notes and summaries from PDFs, EPUBs, or markdown
• watch-video pulls transcripts and key moments from YouTube, Loom, Zoom, and similar sources

Output, operations, and utility skills

• jab-hook plans social content rotation across a portfolio using the "give value, give value, give value, then ask" rhythm
• slide-deck drafts and exports branded React decks
• pm runs a five-column kanban and Eisenhower matrix across multiple projects, with adapters for Notion, GitHub, Linear, and more
• personal-cfo and company-cfo model financial scenarios, from a house purchase to monthly cash flow to a company's cash pulse
• paste cleans terminal output for wherever it's headed next: Slack, LinkedIn, X, email
• social-fetch pulls a social post from any of ten platforms as structured data

The skills are designed to call each other rather than duplicate logic. watch-video alone is referenced by 54 other points in the system, skillify by 51, second-brain by 48. That cross-referencing is the actual design bet here: instead of one skill trying to do everything, each one is narrow and routes to a neighbor when the job isn't really its job.

Where it's different from the usual coding skill

Most of what shows up in this AI Stack lane is developer tooling: coding agents, code review, test generation. Makerskills is closer to a Claude Code skills for founders category of its own, packaging the operator's calendar rather than the engineer's. It's also worth separating from i-have-adhd-claude-code-skill, an earlier pick in this newsletter: that one is about stripping preamble and getting Claude to skip the ceremony in its responses. Makerskills doesn't touch how Claude talks. It's a library of workflows for what to actually do: decide, research, capture, publish, model the numbers.

What to know before installing

The repo is public and generic by design, which means it ships no personal data. Anything specific to your business, portfolio, or voice lives in a separate ~/.config/makerskills/ folder that stays on your machine and out of git. That's the right call for an open-source tool, but it means the first hour is setup, not output: pointing environment variables at your second-brain vault, your CFO report folder, your slide-deck repo.

Some skills lean on paid services to do their real work. domain expects Namecheap and Domainr API access. company-cfo expects bank, payment processor, and payroll data wired in through toolify. company-brain can auto-sync from tools like Fathom or Gong. Without those connections, several skills fall back to a manual, doc-driven mode, which is still useful but not the full pitch.

The frameworks baked in are also opinionated: the 37signals decision method, Gary Vee's content rhythm, a specific domain-hunting methodology. If those don't match how you already think about decisions or content, expect to adapt rather than adopt wholesale. And because the project is only about two months old at time of writing, the finance and CFO skills in particular haven't had much public runway to prove out against real books.

Who this is for

Anyone running their content, research, and business operations through an AI agent already, and looking for a second brain AI workflow that goes past note-taking into actual decisions and output, has reason to look at this. It's free, it's documented well enough to read the SKILL.md files by hand instead of running them, and the scope, an actual operator's job rather than another coding assistant, is genuinely underserved in this space right now.

Repo: github.com/coreyhaines31/makerskills