Bright Data Scraper Studio — Claude Code skill

A skill that teaches Claude Code, Cursor or Codex how to build, run and self-heal custom web scrapers for any public site, entirely from the terminal.

Point your coding agent at a URL, describe the data you want in plain English, and get back a working scraper you own — plus a Collector ID that is already a live API endpoint.

Try Bright Data: brdta.com/iss25 — get $25 of credit to run everything in this skill.

That's my affiliate link. It costs you nothing extra and helps support work like this.

Install

Claude Code

git clone https://github.com/IncomeStreamSurfer/brightdata-scraper-studio-skill.git ~/.claude/skills/brightdata-scraper-studio

opencode

git clone https://github.com/IncomeStreamSurfer/brightdata-scraper-studio-skill.git ~/.config/opencode/skills/brightdata-scraper-studio

Either way it's picked up on the next session — no config, no restart. Same SKILL.md, same frontmatter, both agents read it as-is.

Already have it in Claude Code? Symlink instead of cloning twice, so there's one source of truth to edit:

ln -s ~/.claude/skills/brightdata-scraper-studio ~/.config/opencode/skills/brightdata-scraper-studio

Cursor / Codex: clone anywhere and point your rules file at SKILL.md, or copy its contents into .cursor/rules or AGENTS.md.

Verify it loaded: ask your agent "what do you know about Bright Data Scraper Studio?" — it should reference the five scraper types without searching the web.

What it covers

• Setup in three commands — runs via npx, so no global install, no sudo, no EACCES, no node version manager fights
• Picking the right scraper type — PDP, Discovery, Discovery + PDP, Search, and Sitemap, with the rule for naming the shape explicitly so you don't pay the build time twice
• The build loop — create → run → heal → approve
• Self-healing — when a site changes its HTML, one prompt repairs the scraper you own and keeps the same Collector ID, so every downstream trigger and schedule keeps working
• Pinning the Collector ID in your rules file so agents stop rebuilding the scraper on every session and just run it
• The trigger API — batch and real-time paths, callable from any language or scheduler
• A "check for a faster path first" step — scrape, search, discover and 43 pre-built pipelines all return in seconds, so the agent doesn't burn 15 minutes rebuilding something that already exists

Why the gotchas section matters

Every gotcha in the skill is a failure hit in real use, not a guess:

• zsh globs on unquoted URLs — ...?v=abc dies with no matches found before the CLI runs
• piping a long-running command into head SIGPIPEs the poller and destroys the result
• stale docs examples return {"error": "dead_page"} rather than failing loudly
• scraper create takes 5–25 minutes, so an agent must never start one without telling you

Requirements

• Node.js 20+
• A Bright Data account — brdta.com/iss25 for $25 of credit

No global install needed. The skill runs everything through npx -p @brightdata/cli.

Files

Cost

$1.50 per 1K page loads pay-as-you-go, down to $1.00 at scale. Sign up through brdta.com/iss25 for $25 of credit — that's roughly 16,000 page loads, far more than you need to work through everything here. Run bdata budget to check your balance.

Contributing

Issues and PRs welcome — especially new gotchas found in real use.

License

MIT