Browsecraft

Memory-oriented browser automation skill for repeatable web workflows (login, extraction, bulk actions, form filling, screenshots, checks) across RoxyBrowser, Camoufox, and Chrome.

BrowseCraft Skill

Installation Prerequisites

1. Install CLI: npm install -g browsecraft-cli
2. Verify install: browsecraft --help
3. Optional (RoxyBrowser only): configure your Roxy API endpoint and token in local env/config.

If the CLI is missing, stop and ask the user to install it first. Do not assume local package scripts are available.

Recommended Flow

1. Check status: browsecraft status
2. Start browser if needed: browsecraft start
3. Open target page: browsecraft open <url>
4. Capture snapshot: browsecraft snapshot
5. Prefer click-ref / fill-ref for stable interactions
6. Re-snapshot after page transitions
7. Capture result evidence: browsecraft screenshot

Backend Strategy

• RoxyBrowser: browsecraft start --type roxy --roxy-window-id ...
• Camoufox: browsecraft start --type camoufox (optional --camoufox-path)
• Existing endpoint: browsecraft connect <endpoint> --type <chrome|camoufox|roxy>

Credential & Scope Policy

• Roxy credentials are optional and only needed when the user explicitly chooses RoxyBrowser.
• Prefer environment variables over CLI token flags. Avoid printing or echoing secrets.
• Never send credentials or page data to third-party endpoints unless the user explicitly requests it.
• If a step needs network access outside the target website, ask for confirmation first.

Stability Rules

• If element lookup fails: refresh with snapshot, then retry.
• If page is unstable: use wait-for before interaction.
• Always return structured output: objective / steps / result / failure reason / next action.