A collection of packaged instructions that extend AI coding agent capabilities for browser automation, following the Agent Skills format.
Agent skills for agent-browser - a headless browser automation CLI optimized for AI agents.
| Skill | Description | Compatibility |
|---|---|---|
| agent-browser | Universal skill for browser automation | Claude Code, Codex CLI, OpenCode, Gemini CLI |
| agent-browser-clawdbot | Clawdbot-specific with metadata | Clawdbot |
npx add-skill MaTriXy/agent-browser
# or manually copy to ~/.claude/skills/npx add-skill MaTriXy/agent-browser
# or manually copy to ~/.codex/skills/npx add-skill MaTriXy/agent-browser
# or manually copy to ~/.config/opencode/skills/npx add-skill MaTriXy/agent-browser
# or manually copy to ~/.gemini/skills/Use the agent-browser-clawdbot skill which includes Clawdbot-specific metadata.
A headless browser automation CLI that uses accessibility tree snapshots with ref-based element selection for deterministic actions. Perfect for:
- Multi-step browser automation workflows
- Web scraping with reliable element targeting
- Form filling and submission
- SPA testing with session isolation
Traditional browser automation:
- AI describes what to click
- Code queries DOM for element
- Element might have moved/changed
- Re-query on failure
agent-browser workflow:
- Get snapshot with refs
- AI identifies target ref from snapshot
- Use exact ref (deterministic)
- Fast execution (no re-query)
# Navigate and get snapshot
agent-browser open https://example.com
agent-browser snapshot -i --json
# AI identifies refs from snapshot, then:
agent-browser click @e2 # Click button
agent-browser fill @e3 "text" # Fill inputnpm install -g agent-browser
agent-browser install # Download Chromium
agent-browser install --with-deps # Linux: + system depsSkills created by Yossi Elkrief (@MaTriXy)
agent-browser CLI by Vercel Labs
MIT