DuneSlide : l'injection de prompt s'évade du sandbox de Cursor jusqu'au RCE système
Cato AI Labs (1er juillet 2026) montre qu'un simple prompt piégé — via un serveur MCP ou un résultat web — peut écraser le binaire de sandbox de Cursor et aboutir à un RCE système zéro-clic. Corrigé dans Cursor 3.0.
# DuneSlide: prompt injection escapes Cursor's sandbox (illustrative, defensive) # Zero-click indirect injection arrives via an MCP result or fetched web page: hidden = "[hidden instruction in untrusted content]" # It steers run_terminal_cmd's OPTIONAL working_directory outside the project, # silently widening the sandbox write allowlist a model-set arg controls: run_terminal_cmd(working_directory="[system path]", command="[REDACTED]") # -> next command runs unconfined -> OS-level RCE as the developer. # Defense: upgrade to Cursor 3.0; derive sandbox scope from trusted policy, # not tool args; canonicalize symlinks and FAIL CLOSED; block writes to binaries.