coreyhaines31/makerskills

makerskills – AI‑agent “skill” library for solo founders and indie operators

What it is

• A Claude Code plugin (and compatible with other LLM‑agent hosts like Codex or Cursor) that bundles a curated set of “skills”. Each skill is a small, documented workflow that the agent can invoke via a slash‑command (e.g. /decide, /paste).
• The repository contains the skill definitions (SKILL.md files) and the thin runtime that lets Claude call them, store results, and compose one skill from another.

Why it matters

• Modern AI agents are great at answering questions, but turning those answers into repeatable, structured work‑flows is still hard. makerskills bridges that gap by turning common founder tasks—decision‑making, research, domain‑name hunting, personal finance modeling, knowledge‑base management, content creation, etc.—into plug‑and‑play agent actions.
• Because each skill is documentation‑first (the SKILL.md describes triggers, steps, and version), the same logic can be run manually or automatically, making the system transparent and easy to extend.

Key concepts

Concept	Explanation
Skill	A self‑contained workflow described in a SKILL.md file. It defines a trigger phrase, the steps the agent should take, and where the output is saved.
Plugin install	In Claude Code you add the plugin via the marketplace (/plugin marketplace add coreyhaines31/makerskills) and then install it (/plugin install makerskills@makerskills). A local dev symlink works as well.
Configuration directory	Personal data lives outside the repo under ~/.config/makerskills/ (or a custom path set via MAKERSKILLS_CONFIG). Each skill reads/writes its own sub‑folder there.
Composition	Skills can call other skills by name. For example watch-video may invoke second‑brain to capture a summary, or business‑brainstorm may call deep‑research and domain.
Versioning	The repo has a global semver tag (e.g., v0.5.0). Each skill also carries its own metadata.version inside its front‑matter, allowing independent iteration.

How you use it

1. Install the plugin in Claude Code (or symlink for local development).
2. Set MAKERSKILLS_CONFIG (and optional env vars for vault locations) in your shell.
3. Invoke a skill with a slash command, e.g. /decide. The agent asks for the decision context, runs the 37signals framework, and writes a dated archive file.
4. Browse the corresponding SKILL.md to see the exact steps, triggers, and any required configuration.
5. Mix and match skills—run /business‑brainstorm to pressure‑test an idea, then /domain to hunt a matching .com, or /second‑brain to capture the research into your personal wiki.

The 20 built‑in skills (grouped)

• Meta utilities – skillify (create/adapt/update a skill), toolify (wire up an API or service), loopify (schedule recurring agent loops).
• Decision & strategy – decide, unstuck, maker‑council, business‑brainstorm, domain, deep‑research.
• Knowledge & consumption – second‑brain, company‑brain, read‑book, watch‑video.
• Content creation – jab‑hook (social‑post rotation), slide‑deck.
• Operations & utilities – pm (personal kanban), personal‑cfo, company‑cfo, paste (clean output), social‑fetch.

Architecture snapshot

• The plugin is a thin wrapper around the skill files; the heavy lifting (web searches, API calls, browser automation) is delegated to Claude’s built‑in tools (WebSearch, agent‑browser, etc.) and to external binaries installed via the INSTALL.md instructions.
• All persistent data lives in the user’s config directory, keeping the repository stateless and safe to share.
• Skills declare their dependencies in a ## Composes with section, enabling the agent to resolve call graphs at runtime.

Getting started resources

• INSTALL.md – environment variables, dependency list, local‑dev setup.
• EXAMPLES.md – end‑to‑end walkthroughs for each skill.
• ARCHITECTURE.md – deeper dive into the plugin’s mental model and how skill packs interoperate.
• CONTRIBUTING.md – guidelines for adding new skills or improving existing ones.

Who built it

• Corey Haines, founder of Conversion Factory and Magister, maintains the repo under an MIT license.

Bottom line makerskills turns Claude (and similar LLM agents) into a personal operations platform for founders, letting them automate the repetitive, knowledge‑heavy tasks that normally sit in spreadsheets, notes, and ad‑hoc scripts. By exposing each workflow as a documented, versioned skill, the project stays both transparent and extensible for anyone who wants to tailor their AI‑assistant to their own “maker” workflow.