CPE
PraisonAI resolves tool names against module globals and `__main__` after failing to match declared tools, allowing an attacker who can influence tool-call names to invoke unintended application callables, leading to potential unauthorized state changes and command execution.