Goal Reached Thanks to every supporter — we hit 100%!

Goal: 1000 CNY · Raised: 1359 CNY

100%

Buy Us a Coffee

CVE-2026-47391— PraisonAI's unauthenticated A2A official example can reach real LLM-driven `eval()` tool execution

Quick assessment

Affected

MervinPraison PraisonAI

Exploitation

Public or AI PoC available; prioritize validation

Recommended action

Check the vendor advisory and references for a fixed version. If immediate upgrade is impossible, restrict exposure and increase monitoring.

Mervin Praison PraisonAI是Mervin Praison个人开发者开源的一个低代码多智能体协作框架。 Mervin Praison PraisonAI 4.6.40之前版本存在安全漏洞，该漏洞源于A2A JSON-RPC端点未经身份验证开放，并注册了使用Python eval()实现的calculate工具，可能导致未经身份验证的远程攻击者通过发送message/send请求到/a2a，利用LLM调用calculate工具在服务器进程中执行任意Python代码，并暴露任务历史和任务取

CVSS 9.8 · Critical EPSS 1.17% · P66 Updated Jul 23, 2026

Possible ATT&CK Techniques 3 AI

T1005 · Data from Local System T1059 · Command and Scripting Interpreter T1190 · Exploit Public-Facing Application

Affected Version Matrix 1

Vendor	Product	Version Range	Status
MervinPraison	PraisonAI	`< 4.6.40`	affected

Get alerts for future matching vulnerabilities Log in to subscribe

I. Basic Information for CVE-2026-47391

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Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.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's unauthenticated A2A official example can reach real LLM-driven `eval()` tool execution

Source: CVE Program / CVE List V5

Vulnerability Description

PraisonAI is a multi-agent teams system. Prior to version 4.6.40, PraisonAI's first-party A2A server example exposes an unauthenticated A2A JSON-RPC endpoint and registers a `calculate(expression)` tool implemented with Python `eval()`. The example also binds to `0.0.0.0`. A remote unauthenticated attacker can send `message/send` to `/a2a`; the request reaches `agent.chat()`, and a real LLM can invoke the registered `calculate` tool. In testing with `gemini/gemini-2.5-flash-lite`, this resulted in arbitrary Python execution in the server process, confirmed by creation of a marker file from an unauthenticated HTTP request. The issue affects deployments following the official A2A example or similar unauthenticated public A2A deployments with unsafe tools. The default unauthenticated A2A surface also exposes task history and task cancellation APIs, increasing confidentiality and integrity impact. Version 4.6.40 patches the issue.

Source: CVE Program / CVE List V5

CVSS Information

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

Source: CVE Program / CVE List V5

Vulnerability Type

动态执行代码中指令转义处理不恰当(Eval注入)

Source: CVE Program / CVE List V5

Vulnerability Title

Mervin Praison PraisonAI 授权问题漏洞

Source: CNNVD (China National Vulnerability Database)

Vulnerability Description

Source: CNNVD (China National Vulnerability Database)

CVSS Information

N/A

Source: CNNVD (China National Vulnerability Database)

Vulnerability Type

N/A

Source: CNNVD (China National Vulnerability Database)

Affected Products

Vendor	Product	Affected Versions	CPE	Subscribe
MervinPraison	PraisonAI	< 4.6.40	-

II. Public POCs for CVE-2026-47391

#	POC Description	Source Link	Shenlong Link

AI-Generated POC Premium

Qwen3.6-35B-A3B · 7908 chars

Pro+ exclusive includes:

Vulnerability reproduction recording (real sandbox build + trigger, exclusive)

In-depth vulnerability mechanism

Trigger conditions & impact

Full executable POC code

Exploit chain & mitigation

POC zip download

100+ AI POC generations per month

Upgrade to Pro+ to unlock Limited ¥499/month ¥599 Sign up first

III. Intelligence Information for CVE-2026-47391

请登录查看更多情报信息。

Patches & Fixes for CVE-2026-47391 (1)

🔗 fix: harden JWT secret handling and replace eval() in examples (#1793) · MervinPraison/PraisonAI@e0fb8e7 · GitHub Read full article x_refsource_MISC

Vendor Advisories for CVE-2026-47391 (1)

🔗 Unauthenticated A2A official example can reach real LLM-driven `eval()` tool execution · Advisory · MervinPraison/PraisonAI · GitHub Read full article x_refsource_CONFIRM

https://www.cve.org/CVERecord?id=CVE-2026-47391

Same Patch Batch · MervinPraison · 2026-07-21 · 25 CVEs total

CVE-2026-47392	9.9 CRITICAL	PraisonAI vulnerable to sandbox escape via `print.__self__` builtins module leak in `execu
CVE-2026-47393	9.8 CRITICAL	PraisonAI `deploy --type api` emits a Flask server with authentication disabled by default
CVE-2026-47396	9.8 CRITICAL	PraisonAI call server exposes unauthenticated agent listing, invocation, and deletion when
CVE-2026-47410	9.8 CRITICAL	praisonai-platform: JWT signing key defaults to hardcoded "dev-secret-change-me", allowing
CVE-2026-47413	9.6 CRITICAL	praisonai-platform: Any workspace member can add arbitrary user as owner via POST /workspa
CVE-2026-47416	9.6 CRITICAL	praisonai-platform: Any workspace member can promote themselves (or any other member) to o
CVE-2026-47405	8.8 HIGH	PraisonAI Platform missing role checks let any workspace member become owner and take over
CVE-2026-47399	8.8 HIGH	PraisonAI Platform workspace-scoped routes allow cross-workspace object access by global o
CVE-2026-47415	8.3 HIGH	praisonai-platform: Issue endpoints accept any issue_id without workspace ownership check,
CVE-2026-47419	8.3 HIGH	praisonai-platform: Agent endpoints accept any agent_id without workspace ownership check,
CVE-2026-47409	8.1 HIGH	praisonai-platform: Any workspace member can remove any other member (including the owner)
CVE-2026-47417	8.1 HIGH	praisonai-platform: Comment endpoints accept any issue_id without workspace ownership chec
CVE-2026-47412	8.1 HIGH	praisonai-platform: Any workspace member can delete the entire workspace via DELETE /works
CVE-2026-47418	8.1 HIGH	praisonai-platform: Project endpoints accept any project_id without workspace ownership ch
CVE-2026-47406	8.1 HIGH	praisonai-platform: Dependency endpoints accept any issue_id and dep_id without workspace
CVE-2026-47398	8.1 HIGH	PraisonAI: Arbitrary code execution via unguarded `spec.loader.exec_module` in `agents_gen
CVE-2026-47414	7.6 HIGH	praisonai-platform: Label endpoints accept any label_id and any issue_id without workspace
CVE-2026-47397	7.1 HIGH	PraisonAI has an Arbitrary File Write in Python API
CVE-2026-47408	6.5 MEDIUM	praisonai-platform: list_issue_activity returns activity log for any issue regardless of w
CVE-2026-47411	6.5 MEDIUM	praisonai-platform: Any workspace member can rewrite workspace name, description, and sett

Showing top 20 of 25 CVEs. View all on vendor page → →

IV. Related Vulnerabilities

Same product: PraisonAI

Same vendor: MervinPraison

Same weakness: CWE-95

V. Comments for CVE-2026-47391

No comments yet

Goal Reached Thanks to every supporter — we hit 100%!

CVE-2026-47391— PraisonAI's unauthenticated A2A official example can reach real LLM-driven `eval()` tool execution

Quick assessment

Possible ATT&CK Techniques 3 AI

Affected Version Matrix 1

I. Basic Information for CVE-2026-47391

Vulnerability Information

Vulnerability Title

Vulnerability Description

CVSS Information

Vulnerability Type

Vulnerability Title

Vulnerability Description

CVSS Information

Vulnerability Type

Affected Products

II. Public POCs for CVE-2026-47391

III. Intelligence Information for CVE-2026-47391

Patches & Fixes for CVE-2026-47391 (1)

Vendor Advisories for CVE-2026-47391 (1)

Same Patch Batch · MervinPraison · 2026-07-21 · 25 CVEs total

IV. Related Vulnerabilities

V. Comments for CVE-2026-47391

Leave a comment