Gnarly Skills Stack is a curated monorepo of AI agent skill packs, installer tooling, indexes, and supporting assets for modern coding assistants (Claude Code, Cursor, Codex, Gemini CLI, Antigravity, Copilot, OpenCode, Windsurf, and others). Skills are Markdown-first instruction trees (SKILL.md plus optional scripts, templates, and references) that steer agents toward repeatable, high‑quality workflows.
This repository exists as a portfolio-ready snapshot: one place to study how ecosystems package agent behavior, browse large catalogs offline, and copy individual subtrees into your own ~/.claude/skills/, .cursor/skills/, .agents/skills/, or project-local skill folders.
Portfolio curation and monorepo assembly: John Anthony (@anthonyjohn17).
Upstream authors, trademarks, and per‑subdirectory licenses remain with the original projects. Do not infer from this README that every file is under a single copyright holder; see each subtree’s LICENSE or upstream notice.
| Directory | Summary |
|---|---|
agent-skills/ |
TypeScript monorepo: CLI + marketplace + catalog MCP for installing and managing packaged skills (upstream: Tech Leads Club). |
andrej-karpathy-skills/ |
Practical Claude Code guardrails distilled from Andrej Karpathy’s guidance (single cohesive CLAUDE.md pattern + skills folder). |
antigravity-skills/ |
Very large (1,300+) installable skill library with npm installer (antigravity-awesome-skills), bundles, catalogs, docs, and tooling. Aggregates contributions from many community sources documented in-tree. |
avids-essential-skills/ |
Small daily-use bundles (e.g. Claude Code hygiene audits). |
claude-skills/ |
Broad Claude skill collection including Composio / connect-apps patterns and curated productivity skills. |
claude-video-skill/ |
Watch / video ingestion workflows for Claude and Codex (hooks, scripts, plugins). Fork lineage noted in subtree README (upstream creator). |
cluade-code-skills/ |
Anthropic-style example skills subtree (documents, MCP builder, frontend design, etc.); path name preserves the bundled folder spelling. |
entrepreneur-skills/ |
Early-stage founder workflows (pricing, MVP, marketing plan, validation, …) as Claude plugin–oriented skill folders. |
feynman-skillls/ |
Feynman adjacent tooling (skills + web/docs metadata). Upstream attribution in subtree (Companion / getcompanion-ai); license preserved. |
google-skills/ |
Agent skills aligned with Google Skills / Agent Skills standard covering Gemini, GCP services, and recipes (official issue tracker remains with Google). |
humanizer-skill/ |
Single skill to humanize prose and strip “AI-ish” stylistic fingerprints (upstream: blader/humanizer). |
last30days-skill-main/ |
Research / recap skill asset tree with specs, fixtures, and tests—useful as a structured skill reference implementation. |
mattpocock-skills/ |
Matt Pocock engineering skills newsletter bundle (skills + docs patterns). Original copyright in LICENSE (Matt Pocock). |
new-repo-skills/ |
Sequential forge stages for new repo scaffolding (numbered stages: inventory → melt → recast → polish). |
repo-refactoring-skills/ |
Public-portfolio refactoring skills and design context (Melt → Recast → Polish narrative); documentation-first, human review gated. |
solid-skills/ |
SOLID, TDD, clean code, and design-pattern oriented multi-file skill set for OO codebases (Agent Skills format). |
superpowers-skills/ |
Superpowers development workflow subtree—subagent-driven development, TDD, planning, systematic debugging (obra / Jesse Vincent—license preserved). |
ui-ux-skill/ |
UI/UX guidance subtree based on ui-ux-pro-max-skill patterns. |
write-articles-skills/ |
Article-writing pipeline assets and samples (paired with refactoring / cleanup context folders). |
x-research-skill/ |
X/Twitter research CLI routed via Composio (fork lineage from Rohun Vora’s original; subtree README describes deltas). |
At the repo root you will also find:
AGENTS.md— expansive index linking to hundreds of upstream official and community skills (adapted from the VoltAgent / awesome‑agent‑skills ecosystem framing).SKILLS.md— Claude‑oriented skills index with thematic groupings.
- Treat every skill subtree as untrusted code-adjacent content until you review it for your threat model (scripts, hooks, network calls, and prompt injection surfaces).
- Prefer installing only the skills you need; avoid wholesale copy of huge trees into production CI without pinning and audit.
- The
agent-skills/subtree documents upstream security expectations inSECURITY.md.
Host-specific global paths (vary by product version):
| Tool | Typical project path | Typical global path |
|---|---|---|
| Cursor | .cursor/skills/ |
~/.cursor/skills/ |
| Claude Code | .claude/skills/ |
~/.claude/skills/ |
| Codex | .agents/skills/ |
~/.agents/skills/ |
| Antigravity | .agent/skills/ |
~/.gemini/antigravity/skills/ |
| Gemini CLI | .gemini/skills/ |
~/.gemini/skills/ |
Many subtrees ship their own installer (npx, skills.sh, or npm scripts)—read each README for authoritative install flags.
git clone https://github.com/anthonyjohn17/gnarly-skills-stack.git
cd gnarly-skills-stackBrowse any subdirectory; symlink or copy only the packs you intend to use into your assistant’s configured skills directory.
If you cloned from a tarball or created this tree locally without a remote yet:
git remote add origin https://github.com/anthonyjohn17/gnarly-skills-stack.git
git branch -M main
git push -u origin mainSee LICENSE at the repo root for the licensing note on aggregated documentation.
Individual subtrees retain their upstream licenses (MIT, Apache-2.0, ISC, and others). Removing or rewriting third-party copyright lines from upstream LICENSE files would violate those terms—those files stay intact in this portfolio copy.
This monorepo stands on enormous community and vendor work—Anthropic’s Agent Skills framing, VoltAgent’s curation hubs, thousands of anonymous contributors across bundled catalogs, and the maintainers credited in each subtree’s README and upstream links. Maintained as a portfolio collection by anthonyjohn17.