CVE-2026-57120— PraisonAI: execute_code sandbox bypass: str.format C-level attribute access reads every blocklisted dunder
Quick assessment
Affected
MervinPraison praisonaiagents
Exploitation
No confirmed in-the-wild exploitation; assess based on exposure
Recommended action
Check the vendor advisory and references for a fixed version. If immediate upgrade is impossible, restrict exposure and increase monitoring.
PraisonAI 是一个多智能体团队系统。在 praisonaiagents 1.6.59 版本之前， 的沙箱模式允许在运行时组装被屏蔽的 dunder（双下划线）名称，并允许通过 C 层级的属性访问让 或 解析带点的字段，从而绕过 。当审批被自动授予时，这将使类名、限定名、基类、全局变量以及对象字典等属性暴露给受提示词（prompt）影响的代码，从而形成一种高影响的读取原语（read primitive），但未构成完整的进程内执行链。该问题已在 praisonaiagents 1.6.59 版本中修复。
Possible ATT&CK Techniques 1 AI
T1059 · Command and Scripting Interpreter
Get alerts for future matching vulnerabilities
Log in to subscribe
I. Basic Information for CVE-2026-57120
Vulnerability Information
Have questions about the vulnerability? See if Shenlong's analysis helps!
View Shenlong Deep Dive ↗
Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.
Vulnerability Title
PraisonAI: execute_code sandbox bypass: str.format C-level attribute access reads every blocklisted dunder
Source: CVE Program / CVE List V5
Vulnerability Description
PraisonAI is a multi-agent teams system. Prior to praisonaiagents 1.6.59, execute_code sandbox mode permits runtime assembly of blocklisted dunder names and allows str.format or str.format_map to resolve dotted fields through C-level attribute access that bypasses _safe_getattr. This exposes class, qualified-name, base-class, globals, and object-dictionary attributes to prompt-influenced code when approval is automatically granted, producing a high-impact read primitive without establishing a complete in-process execution chain. This issue is fixed in praisonaiagents 1.6.59.
Source: CVE Program / CVE List V5
CVSS Information
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
Source: CVE Program / CVE List V5
Vulnerability Type
保护机制失效
Source: CVE Program / CVE List V5
Affected Products
| Vendor | Product | Affected Versions | CPE | Subscribe |
|---|---|---|---|---|
| MervinPraison | praisonaiagents | < 1.6.59 | - |
|
II. Public POCs for CVE-2026-57120
| # | POC Description | Source Link | Shenlong Link |
|---|
AI-Generated POC
Premium
No public POC found.
Login to generate AI POCIII. Intelligence Information for CVE-2026-57120
请登录查看更多情报信息。
Patches & Fixes for CVE-2026-57120 (1)
Vendor Advisories for CVE-2026-57120 (1)
Vendor Pages for CVE-2026-57120 (1)
-
🔗 Release PraisonAI v4.6.59 · MervinPraison/PraisonAI · GitHub x_refsource_MISC
Same Patch Batch · MervinPraison · 2026-09-14 · 16 CVEs total
| CVE-2026-57123 | 9.8 CRITICAL | PraisonAI: MCP SSE transport binds 0.0.0.0 with no authentication and no Origin validation |
| CVE-2026-57124 | 9.8 CRITICAL | PraisonAI UI MCP connect endpoint allows unauthenticated local command execution |
| CVE-2026-57125 | 9.8 CRITICAL | PraisonAI: Unauthenticated RCE via Jobs API + Approval Bypass |
| CVE-2026-57127 | 9.8 CRITICAL | praisonai: recipe serve auth middleware silently disables itself when no secret is set |
| CVE-2026-57131 | 9.8 CRITICAL | praisonai: Jobs API exposes agent-execution endpoints with no authentication |
| CVE-2026-57145 | 9.1 CRITICAL | PraisonAI: Arbitrary File Read/Write via `multiedit` Tool Without Path Validation |
| CVE-2026-57122 | 8.6 HIGH | PraisonAI: Webhook signature verification skipped (fail-open) when secret unset, allowing |
| CVE-2026-57126 | 8.5 HIGH | praisonaiagents: SSRF guard validates literal IPs only and never resolves DNS |
| CVE-2026-57132 | 8.2 HIGH | PraisonAI: PRAISONAI_CALL_AUTH=disabled environment variable unconditionally disables auth |
| CVE-2026-57130 | 8.1 HIGH | PraisonAI: IMAP Command Injection via Unsanitized Email Search Parameters |
| CVE-2026-57119 | 7.5 HIGH | PraisonAI: Unauthenticated Local File Inclusion via agent_file path in the Jobs API |
| CVE-2026-57129 | 7.5 HIGH | PraisonAI: Arbitrary File Read via `@file:` Mention Path Traversal |
| CVE-2026-56839 | 7.3 HIGH | PraisonAI Code agent tools fail open without a workspace boundary |
| CVE-2026-57115 | 6.5 MEDIUM | PraisonAI: SpiderTools redirect-target SSRF protection bypass |
| CVE-2026-57128 | 4.3 MEDIUM | PraisonAI: Unauthenticated Event Injection via SSE `/publish` Endpoint |
IV. Related Vulnerabilities
Same product: praisonaiagents
-
CVE-2026-57129
PraisonAI: Arbitrary File Read via `@file:` Mention Path Tra -
CVE-2026-57123
PraisonAI: MCP SSE transport binds 0.0.0.0 with no authentic -
CVE-2026-57130
PraisonAI: IMAP Command Injection via Unsanitized Email Sear -
CVE-2026-40160
PraisonAIAgents has SSRF via unvalidated URL in `web_crawl` -
CVE-2026-40153
PraisonAIAgents Affected by Environment Variable Secret Exfi
Same vendor: MervinPraison
-
CVE-2026-57145
PraisonAI: Arbitrary File Read/Write via `multiedit` Tool Wi -
CVE-2026-57132
PraisonAI: PRAISONAI_CALL_AUTH=disabled environment variable -
CVE-2026-57131
praisonai: Jobs API exposes agent-execution endpoints with n -
CVE-2026-57124
PraisonAI UI MCP connect endpoint allows unauthenticated loc -
CVE-2026-57122
PraisonAI: Webhook signature verification skipped (fail-open
Same weakness: CWE-693
-
CVE-2026-76059
Langflow is vulnerable to arbitrary code execution due to mu -
CVE-2026-45770
Suricata lua: excessive flow variable registration can bypas -
CVE-2026-0306
Prisma Access Agent: EndPoint DLP Bypass Vulnerability on Wi -
CVE-2026-79638
Dell Secure Connect Gateway 处理逻辑错误漏洞 -
CVE-2026-87808
SiYuan before v3.8.2 Read-Only Boundary Bypass via fullTextS
V. Comments for CVE-2026-57120
No comments yet