Security context

What an agent needs to avoid regressing past fixes and find the next vuln in this repo.

labring/fastgpt
main @ aa1d879
75
Fixes
14
CVEs
CRITICAL
Peak severity
20.0%
Coverage
Highlights
Auth Bypass: 39 prior fixes. Scrutinize any change in this area.
packages/service/common/system/utils.ts: most-fixed (3 issues). Treat as high-risk during review.
50 high-severity fixes in this history; regressions here are high-impact.
Recurring patterns

The bug types that recur here, drawn from past fixes, not open vulnerabilities.

Sandbox Escape: The platform relies on executing untrusted code within sandboxes. Weak isolation configurations (such as permissive syscalls and weak standard output redirection) historically allowed complete sandbox escape and host file system manipulation. Hardening required migrating to one-shot chroot/seccomp structures and strict AST verification.
Auth Bypass: Critical routing and certificate parser failures (such as empty token secrets or permissive fallbacks in parseHeaderCert) allowed attackers to bypass authentication entirely, accessing administrative resources and other tenants' collections.
SSRF: Outbound requests driven by workflows, MCP tools, and HTTP integrations frequently bypassed internal IP checks, exposing cloud metadata and intranet endpoints to remote attackers. Fixes introduced deep subnet resolving and DNS validation using ipaddr.js.