CriticalPublic exploit

Sandbox escape in PraisonAI praisonai-agents execute_code()

IdentifiersCVE-2026-34938CWE-693· Protection Mechanism Failure

CVE-2026-34938 affects PraisonAI praisonai-agents prior to version 1.5.90. The flaw is in the execute_code() functionality, which executes attacker-controlled Python code inside a purported three-layer sandbox. That sandbox can be fully bypassed by supplying a str subclass with an overridden startswith() method to the _safe_getattr wrapper. The weakness allows sandbox restrictions to be evaded and attacker-controlled Python to reach dangerous functionality, resulting in arbitrary OS command execution on the host running the vulnerable software.

For your environment

Are you exposed to this one?

Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.

Start free trial

ANALYST BRIEF

Impact, mitigation & remediation

What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation results in a complete sandbox escape and arbitrary operating system command execution in the security context of the affected process. This can enable compromise of confidentiality, integrity, and availability, including reading and modifying files accessible to the process, theft of environment variables and credentials, execution of follow-on payloads, persistence, and potential pivoting into adjacent internal systems reachable from the host.

Mitigation

If you can’t patch tonight, do this now.

If immediate upgrade is not possible, restrict or disable untrusted use of execute_code() and prevent attacker-controlled Python from reaching the vulnerable sandbox path. Additional hardening measures described in the provided content include strengthening attribute filtering in _safe_getattr and related denylist logic, but these should be treated only as temporary risk reduction. Do not rely on denylist-based in-process Python sandboxing as a primary security boundary; use stronger external isolation controls instead.

Remediation

Patch, then assume compromise.

Upgrade praisonai-agents to version 1.5.90 or later, which patches the sandbox bypass. Because the issue stems from bypass of in-process Python sandbox controls, remediation should prioritize the vendor fix and, where feasible, stronger isolation for untrusted code execution such as OS-level sandboxing, containers, or separate execution environments.

PUBLIC EXPLOITS

Exploits

No public exploits tracked yet. Mallory keeps watching.

VALID 0 / 0 TOTALView more in app

No public exploit code observed for this vulnerability.

EXPOSURE SURFACE

Affected products & vendors

Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.

VendorProductType

PraisonPraisonaiagentsapplication

PraisonAIPraisonaiagentsapplication

Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.

ACTIVITY FEED

Recent activity

14 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

14 SOURCESView more in app

cvefeed high severityNews

Apr 3, 2026

CVE-2026-34938 - PraisonAI: Python Sandbox Escape via str Subclass startswith() Override in execute_code

A sandbox bypass vulnerability in PraisonAI praisonai-agents execute_code() that allows attacker-controlled Python to achieve arbitrary OS command execution on the host.

What this page doesn’t show

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, which detections to deploy, and what to do tonight.

Start free trial

Exposure mapping

Query your assets running an affected version, and investigate the blast radius.

Threat actor evidence

Every observed campaign linking this CVE to a named adversary.

Associated malware

Malware families riding this exploit, with evidence and IOCs.

Detection signatures

YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.

Vendor-by-vendor mapping

Cross-references every affected SKU, including bundled OEM variants.

Social activity13