Security context

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

agent0ai/agent-zero

              main @ 6a6cecf
            

Fixes

CVEs

CRITICAL

Peak severity

37.1%

Coverage

Highlights

Path Traversal: 12 prior fixes. Scrutinize any change in this area.

webui/js/api.js: most-fixed (3 issues). Treat as high-risk during review.

30 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.

Path Traversal: Most frequently patched class; multiple endpoints (download, image, plugin, skill, extension asset) lacked containment checks, allowing arbitrary file read/write/delete. Fixes added is_in_dir/safe_filename but some (d541a94, 08693c2) expanded allowed paths, indicating ongoing boundary tuning.

Auth Bypass: Ten distinct bypasses across policy enforcement, origin checks, race conditions, and stale state. Fixes involved canonicalization, scope checks, locking, and resetting flags—indicating systemic gaps in authorization validation across contexts and APIs.

Information Disclosure: Secret masking gaps in summaries, command output, tool responses, and streams. Multiple fixes added mask_values to different sinks, suggesting incomplete coverage across the agent's output paths.