Skills Kit

An opinionated collection of Agent Skills I use daily in my work.

Not a marketplace—a toolkit. These are skills I've built and curated for my own workflows with Claude Code. If they're useful to you too, great.

Quick Start

# Add the kit
/plugin marketplace add nibzard/skills

# Install a skill
/plugin install yt-transcript@skills

The Skills

Data & Notebooks

Skill What it does
marimo Reactive Python notebooks for data analysis, dashboards, and interactive apps
steel-org-recon Reconstruct what a Steel customer org was doing in its browser sessions and whether it succeeded, from warehouse records plus raw session logs

Media & Content

Skill What it does
yt-transcript Fetch YouTube video transcripts for analysis and summarization
brand-illustrator Generate Builder Methods-style line art illustrations with accent colors

Productivity

Skill What it does
claude-thread-publisher Publish Claude Code conversations as shareable HTML via GitHub Gists
tmux Test interactive CLI/TUI applications end-to-end using tmux sessions

Development

Skill What it does
release-runbook Universal release workflow: detect project type, test, bump, tag, push
skill-creator Create new Agent Skills from templates or interactively

Security

Skill What it does
pentest-toolkit Authorized security testing and vulnerability assessment tools

Why This Exists

Agent Skills are how Claude Code extends its capabilities—model-invoked, context-aware, and progressively loaded. I built this collection because:

  1. I use them—every skill here solves a real problem I encounter
  2. They work—each is tested in production workflows
  3. They're opinionated—built my way, for my workflows

If you're looking for a public marketplace with community contributions, this isn't it. If you want tools that have been battle-tested in real work, read on.

Installing

Install the Kit

/plugin marketplace add nibzard/skills

Install Individual Skills

# YouTube transcripts
/plugin install yt-transcript@skills

# Publish conversations
/plugin install claude-thread-publisher@skills

# Run releases
/plugin install release-runbook@skills

Install Everything

// .claude/settings.json
{
  "enabledPlugins": [
    "marimo@skills",
    "yt-transcript@skills",
    "claude-thread-publisher@skills",
    "tmux@skills",
    "release-runbook@skills"
  ]
}

Installing with skillctl

skillctl is a local-first skill manager that projects skills into the roots multiple agent runtimes actually read (Claude Code, Codex, Gemini CLI, Copilot, Amp, OpenCode). One tool, every runtime.

Install skillctl

curl -fsSL https://github.com/nibzard/skillctl/releases/latest/download/skillctl-install.sh | sh

Install a skill from this repo

# Initialize skillctl in your workspace
skillctl init

# Install a skill
skillctl install https://github.com/nibzard/skills --name yt-transcript --subpath skills/yt-transcript

# Project into runtime roots
skillctl sync

Manage installed skills

skillctl list              # what's installed and where it projects
skillctl update yt-transcript   # pull upstream changes
skillctl explain yt-transcript  # why a skill is active or shadowed
skillctl doctor            # diagnose drift, conflicts, missing projections
skillctl override yt-transcript # customize without forking
skillctl fork yt-transcript    # take full local ownership
skillctl remove yt-transcript  # uninstall

See the skillctl README for the full command reference.

Contributing

This is a personal toolkit, but I accept quality PRs that align with the collection's scope:

  • New skills must solve a real, generalizable problem
  • Improvements should maintain existing functionality
  • Documentation must be clear and complete

Before contributing, check that your skill isn't better suited as a standalone plugin or separate marketplace.

Structure

skills/
├── .claude-plugin/
│   └── marketplace.json       # Kit manifest
├── skills/
│   ├── marimo/                # Reactive notebooks
│   ├── yt-transcript/         # YouTube transcripts
│   ├── claude-thread-publisher/  # Share conversations
│   ├── tmux/                  # TUI testing
│   ├── release-runbook/       # Release automation
│   ├── skill-creator/         # Skill scaffolding
│   ├── pentest-toolkit/       # Security tools
│   ├── brand-illustrator/     # Illustration generation
│   └── steel-org-recon/       # Steel session forensics & outcomes
├── docs/                      # Additional docs
└── README.md

About Agent Skills

Agent Skills extend Claude Code through organized folders containing instructions, scripts, and resources. Unlike traditional commands, Skills are model-invoked—Claude decides when to use them based on your request and the skill's description.

Key characteristics:

  • Autonomous—Claude selects skills automatically based on context
  • Modular—Each skill is self-contained with metadata
  • Progressive—Load additional files only when referenced
  • Permission-aware—Fine-grained tool access control

Documentation

License

MIT — see LICENSE. Individual skills may have their own licenses.

nibzard · Issues · Discussion