Tool Policy Bypass in PraisonAI Multi-Agent Systems by Mervin Praison
CVE-2026-56838

7.8HIGH

Key Information:

Vendor

Mervinpraison

Status
Praisonai
Vendor
CVE Published:
23 October 2026

What is CVE-2026-56838?

PraisonAI, a multi-agent team system, has a vulnerability where the recipe execution process bypasses checks for dangerous tools. Specifically, while the recipe.run() function accurately enforces tool policies, the streaming approach (recipe.run_stream()) neglects to apply these checks. This flaw allows users to execute recipes that include potentially harmful commands, evading the security measures designed to protect against the use of dangerous tools. The vulnerability is present in versions 4.5.87 to before 4.6.59 and is addressed in the latter version.

Affected Version(s)

PraisonAI >= 4.5.87, < 4.6.59

References

https://github.com/MervinPraison/PraisonAI/security/advis...
x_refsource_CONFIRM

CVSS V3.1

Score:
7.8
Severity:
HIGH
Confidentiality:
High
Integrity:
High
Availability:
High
Attack Vector:
Local
Attack Complexity:
Low
Privileges Required:
None
User Interaction:
Required
Scope:
Unchanged

Timeline

  • Vulnerability published

  • Vulnerability Reserved

.