Security context

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

arize-ai/phoenix

              main @ 6e93f93
            

Fixes

CVEs

CRITICAL

Peak severity

33.3%

Coverage

Highlights

Auth Bypass: 21 prior fixes. Scrutinize any change in this area.

GitHub: most-fixed (4 issues). Treat as high-risk during review.

29 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.

Auth Bypass: The application struggled with correctly isolating distinct token classifications, allowing refresh tokens to be parsed and accepted as access tokens for general HTTP/gRPC requests. Authentication validation must strictly assert the token's claim class.

SSRF: Playground client endpoints allowed server-configured credentials (like API keys) to be leaked to external URLs by accepting arbitrary client-supplied base URLs during fallback operations. Custom base URLs must be explicitly rejected if fallback environment credentials are used.

Command Injection: Privileged agent operations executed code or attached bash tools without administrative restrictions. These dangerous behaviors must be gated strictly by environment configuration flags.