Skill Creator

Purpose

Help users create well-structured skills documenting how to scrape specific websites. New skills are saved to the skills/ directory and become immediately usable after an agent restart.

Workflow

Step 1 — Gather requirements

Ask the user:

• What website do they want to scrape? (get the exact URL)
• What specific data fields do they need?
• Any special requirements? (login, pagination, JS rendering, rate limits)

Step 2 — Investigate the website (if needed)

• Use fetch to retrieve the page HTML
• If the page appears empty, switch to playwright
• Identify exact CSS selectors for each required data field
• Note whether JS is required (use playwright vs fetch)

Step 3 — Generate the skill name

• Kebab-case, all lowercase, e.g.: , ,