Agent Skills

Portable agent skills for specialized tasks — image generation, SEO, research, content, engineering workflow. Works with any skills-compatible coding agent: Claude Code, OpenAI Codex, Cursor, Windsurf, opencode, and others.

Skills extend your agent automatically based on context. Describe what you need; the agent picks the right skill.

Prerequisites

uv (Python package manager)

Python-based skills use uv for inline-dependency scripts.

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or via Homebrew
brew install uv

Node (for the skills installer)

npx ships with Node — install Node 18+ from nodejs.org or via brew install node if you don't already have it.

Installation

Use npx skills (by Vercel Labs) to install from this repo.

Preview available skills

npx skills add nc9/skills --list

Install to the current project

# one skill
npx skills add nc9/skills --skill generate-image

# multiple skills
npx skills add nc9/skills --skill generate-image --skill remove-background --skill google-search

Install globally (all projects)

npx skills add nc9/skills --skill generate-image -g

Target specific agents

npx skills auto-detects installed agents, or use -a to pick explicitly:

# Claude Code + Codex + Cursor in one shot
npx skills add nc9/skills --skill generate-image -a claude-code -a codex -a cursor

Install everything, non-interactively (CI/CD)

npx skills add nc9/skills --all -g -y

Other useful commands

npx skills list                       # show installed
npx skills update                     # pull latest
npx skills remove generate-image      # uninstall
npx skills find <keyword>             # search

Install a single skill from a subpath URL:

npx skills add https://github.com/nc9/skills/tree/main/generate-image

Skills

Image & design

Skill	Description	API
generate-image	Image generation & editing (OpenAI gpt-image-2 default, Gemini 3.1 Flash alt)	OpenAI + OpenRouter
remove-background	AI background removal using BiRefNet	None (local)
optimize-image-web	WebP conversion, favicons, social cards, thumbnails	None (local)
generate-favicon	Favicon / icon set generation	None (local)

Search & research

Skill	Description	API
google-search	Google search — web, news, images, videos, places, shopping, scholar, patents	Serper
parallel-web-search	Agentic web search with citations	Parallel AI
parallel-deep-research	Deep research reports with multi-source synthesis	Parallel AI
wayback	Search & fetch archived web pages	Wayback Machine (free)
browserbase	Fetch pages through proxy + captcha solving	Browserbase

SEO & indexing

Skill	Description	API
keyword-research	Keyword ideas, volume, CPC, competition	DataForSEO
google-indexing	Submit / remove / check URL indexing	Google Cloud (free)
google-search-console	Search analytics, URL inspection, sitemaps	Google Cloud (free)

Writing & content

Skill	Description	API
write-content	Blog posts, articles, content pieces with research + humanization	—
humanizer	Remove signs of AI writing — 65 patterns, narrative-first strategy	None (local)
ai-writing-detector	Detect AI-generated text / plagiarism	Pangram

Engineering workflow

Skill	Description	API
commit	Pre-commit workflow — test, lint, format, type-check, atomic commits	None (local)
review	Structured code review via OpenAI Codex MCP — 0-5 score, prioritized issues with file:line citations, verification pass; context from GitHub/Linear/Sentry + plans + conversation	OpenAI (Codex MCP or CLI)
web-ui	Principles for building great web app interfaces	None (local)

Configuration

Skills requiring API keys read from .env or environment variables. Set them per project or globally:

# generate-image (OpenAI SDK direct for gpt-image-2)
OPENAI_API_KEY=sk-...
OPENROUTER_API_KEY=sk-or-...          # only if using Gemini via OpenRouter

# google-search
SERPER_API_KEY=...

# parallel-web-search, parallel-deep-research
PARALLEL_API_KEY=...

# keyword-research
DATAFORSEO_USERNAME=your_email
DATAFORSEO_PASSWORD=your_api_password

# google-indexing, google-search-console
GOOGLE_INDEXING_KEY_FILE=/path/to/service-account-key.json

# ai-writing-detector
PANGRAM_API_KEY=...

# browserbase
BROWSERBASE_API_KEY=...
BROWSERBASE_PROJECT_ID=...

# review
OPENAI_API_KEY=sk-...                 # shared with generate-image

See each skill's SKILL.md for detailed setup.

Usage

After install, describe what you need:

• "Generate an app icon for my SaaS, purple corner badge"
• "Remove the background from this photo"
• "Find keyword ideas for 'python tutorial'"
• "Search Google for recent AI regulation news"
• "Research the competitive landscape of cloud providers"
• "Fetch the archived version of example.com from 2023"
• "Submit these URLs to Google for indexing"
• "Show top search queries for my site"
• "Write a blog post about WebGPU, humanize the output"
• "Review my changes before I commit"

The agent picks the right skill automatically.

Development

Cloning this repo to hack on skills? Use the Makefile instead of npx skills — it symlinks every skill with a SKILL.md into ~/.agents/skills/ (generic) and ~/.claude/skills/ (Claude Code), so edits are picked up live with no reinstall:

git clone https://github.com/nc9/skills.git && cd skills
make link

The target auto-discovers skills via find, so new skills are linked automatically on the next make link. To add a new skill, create a directory with a SKILL.md manifest (see the spec) and run make link again.

Links

• Skills Specification
• npx skills installer
• Agents: Claude Code, Codex, Cursor, opencode, Windsurf