Security context

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

vanna-ai/vanna

              main @ 365d061
            

Fixes

CVEs

CRITICAL

Peak severity

0.0%

Coverage

Highlights

Auth Bypass: 2 prior fixes. Scrutinize any change in this area.

src/vanna/base/base.py: most-fixed (2 issues). Treat as high-risk during review.

3 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.

SQL Injection: The database query execution interface is highly vulnerable to SQL injection because unsanitized commands, particularly those containing non-SELECT statements, can be executed directly. Mitigating this risk requires strict validation of input queries to ensure only safe, read-only SELECT statements are allowed.

Auth Bypass: Local tools registered in the registry can be executed without proper authorization. Without group-membership or access control list verification, untrusted users can trigger restricted administrative actions.

Auth Bypass: Improper parameter propagation allows LLMs to retrieve or interact with restricted data. Forcing explicit parameters like allow_llm_to_see_data is critical to preventing authorization bypasses during SQL generation.