Please do not open a public issue for security problems.

Report privately through GitHub Security Advisories, which lets us work on a fix before disclosure.

Expect an acknowledgement within 72 hours and an assessment within 7 days. If a fix is warranted we'll agree a disclosure timeline with you, and credit you in the advisory and release notes unless you'd rather we didn't.

The latest release on main. This project does not currently backport fixes.

DeepScrape fetches URLs supplied by its operator and executes browser automation, so the interesting attack surface is anything that crosses that boundary:

In scope

• SSRF — bypassing the DNS-resolving guard on any fetch or redirect hop, including reaching private ranges or cloud metadata endpoints
• Authentication or authorisation bypass on API keys, rate limits, or per-key quotas
• Injection through scraped content into extraction output, stored sessions, or MCP tool definitions
• Path traversal via export, ZIP download, or cache paths
• Sandbox escape from the Playwright browser context
• Secrets leaking into logs, error responses, or cached artefacts
• Denial of service through crawl limits, ReDoS in include/exclude patterns, or queue exhaustion

Out of scope

• Anything requiring an already-compromised host or a malicious operator — DeepScrape trusts whoever runs it
• Missing hardening on a deployment the operator chose to expose without an API key
• Rate limiting on a self-hosted instance the reporter controls
• Vulnerabilities in target websites reached through DeepScrape
• Automated scanner output with no demonstrated impact

This is a scraping tool. It respects robots.txt and does not solve CAPTCHAs, and we'd ask that you don't use it to gather data you don't have permission to gather.