Rick Hightower's Universal Skill Installer

Table of content

Rick Hightower is an enterprise architect and AI engineer based in Austin, Texas. He created skilz, the universal package manager for AI agent skills, and co-founded SkillzWave, a marketplace hosting over 44,000 agent skills.

His approach treats AI skills like software packages. Think pip install or npm install, but for AI coding assistants.

Background

Led ML/AI development at Capital One, improving accuracy from 70% to 90% on customer experience metrics
Built systems at Apple, Capital One, and the NFL over 20+ years
Holds TensorFlow certification and Stanford ML Specialization
Multiple Anthropic certifications: Claude SDK (Vertex AI, Bedrock), Claude Tools, MCP

GitHub | Twitter | Medium | LinkedIn

The Problem with Fragmented Skills

Every AI coding assistant has its own skill format. Claude Code uses SKILL.md files. Cursor has its own system. Codex, Gemini, Copilot CLI all work differently. If you write a skill for one, you rewrite it for others.

Hightower’s insight: skills are just markdown instructions with metadata. The differences are surface-level. A universal installer can translate between formats.

skilz: The Universal Package Manager

skilz is a CLI that installs agent skills from any Git repository across 14+ coding assistants:

# Install from GitHub
skilz install SpillwaveSolutions/project-memory

# List installed skills
skilz list

# Update all skills
skilz update

# Search for skills
skilz search typescript

Supported platforms:

Platform	Type
Claude Code	Native
Cursor	Native
OpenCode	Native
Codex	Native
Gemini CLI	Native
GitHub Copilot CLI	Native
Aider	Native
Windsurf	Universal
Qwen Code	Universal
Kimi Code	Universal
Plandex	Universal
Zed AI	Universal
Claude Desktop	Universal

Universal installs work by placing skills in .skilz/skills/ and updating a shared Markdown config file that agents can read.

SkillzWave Marketplace

SkillzWave is the marketplace side of the equation. It hosts 44,000+ skills organized by category:

Development: 16,000+ skills for code review, testing, documentation
Legal: Contract analysis, compliance checking
Finance: DAX queries, financial modeling
DevOps: Deployment pipelines, infrastructure

Skills follow the Agent Skills Standard, which means they work across platforms.

Writing Good Skills

From Hightower’s Claude Code Skills Deep Dive series:

An agent skill is a folder with a SKILL.md file containing:

YAML frontmatter: metadata telling the agent when to activate
Markdown body: instructions the agent follows when active

---
name: project-memory
description: Track project state across sessions
triggers:
  - "remember this"
  - "what was I working on"
---

## Instructions

1. Load existing memory from .project-memory/
2. Update with current context
3. Save state before session ends

Hightower emphasizes: “Skills should be orchestrators that load what they need, not encyclopedias that carry everything.”

Quality Grading

SkillzWave grades skills on a 100-point scale. Hightower’s rigorous reviews exposed problems in the ecosystem:

Files listed in READMEs that didn’t exist
Stub scripts with placeholder code
Generic boilerplate masquerading as documentation

His feedback drove thousands of lines of fixes across major skill repositories, including Jesse Vincent’s Superpowers framework.

Key Takeaways

Principle	Implementation
Skills are portable	One skill, many platforms
Package managers work	`skilz install` beats manual setup
Quality matters	Grade and validate skills
Orchestrate, don’t encyclopedize	Load context on demand

Links

Next: Jesse Vincent’s Superpowers