Universal skills for AI coding agents. Each skill is a self-contained directory you can drop into your local setup for Claude Code, Gemini CLI, or Codex CLI.
| Platform | Install Path | Format |
|---|---|---|
| Claude Code | ~/.claude/skills/<skill>/ |
SKILL.md |
| Gemini CLI | ~/.gemini/skills/<skill>/ |
SKILL.md |
| Codex CLI | Project root or ~/.codex/ |
AGENTS.md |
| Skill | Description | Platforms | Tags |
|---|---|---|---|
| playwright-autopilot | Use when user asks to "automate" a browser task, "write a playwright script", or explicitly mentions playwright automation. Do NOT trigger on general web scraping, testing, or form-filling mentions unless playwright/automation is explicitly referenced. | claude-code, gemini-cli, codex-cli | browser automation playwright scraping mcp |
Your AI agent explores a live browser and hands you the Python script that reproduces every action.
Most browser automation starts with writing a script and hoping it works. Playwright Autopilot flips this: the agent opens a real browser via MCP tools, interacts with pages step by step, and translates each action into Python code as it goes — building your script incrementally through live exploration.
graph LR
A[Navigate] --> B[Observe Page]
B --> C[Act via MCP]
C --> D[Verify Screenshot]
D --> E[Translate to Python]
E --> B
Why it's different:
- No upfront scripting — the agent explores the live page first, writes code second
- Trial-and-error recovery — failed selectors get automatic retries with alternatives, CAPTCHAs pause for human help
- Production-grade output — every script is a class with CLI args, logging, error handling, and accessible selectors
Clone this repo and copy the pre-built skill for your platform:
git clone https://github.com/aghaawais/skill-factory.gitClaude Code:
cp -r skill-factory/dist/claude-code/<skill-name> ~/.claude/skills/Gemini CLI:
cp -r skill-factory/dist/gemini-cli/<skill-name> ~/.gemini/skills/Codex CLI:
cp skill-factory/dist/codex-cli/<skill-name>/AGENTS.md ./AGENTS.mdIf your platform uses SKILL.md (Claude Code, Gemini CLI):
cp -r skill-factory/skills/<skill-name> ~/.claude/skills/
# or
cp -r skill-factory/skills/<skill-name> ~/.gemini/skills/skills/<skill-name>/
├── SKILL.md # Skill definition with YAML frontmatter (source of truth)
└── evals/
└── evals.json # Evaluation test cases
---
name: skill-name # Required — skill identifier
description: When to trigger # Required — activation criteria
version: 1.0.0 # Optional — semver
tags: [tag1, tag2] # Optional — for catalog filtering
platforms: [claude-code, gemini-cli, codex-cli] # Optional — target platforms
author: github-username # Optional — contributor attribution
---See CONTRIBUTING.md for the full guide.
Quick start:
- Create
skills/<your-skill>/SKILL.mdwith frontmatter - Add
evals/evals.jsonwith test cases - Run
npm run buildto generate dist/ files - Submit a PR
MIT