Preliminary research

Configuration-Based Sandbox Escape in AI Coding Tools

Ilan Kalendarov, Ben Zamir and Elad Beber · Cymulate

AI-collected research leads through 2 October 2026, including bounded month-by-month reviews of selected social and community sources from March through September. Unranked, incomplete, not community-vetted, and subject to change.

Finds recurring sandbox escapes in AI coding tools where project-controlled configuration remains writable or is trusted after a restart. The case studies turn settings files, hooks and executable-resolution behavior into host code execution across local and cloud agent environments without breaking the container at the operating-system layer.

Record

Researcher

Ilan Kalendarov, Ben Zamir and Elad Beber

Published by

Cymulate

In the archive

• Collected in the 2026 preliminary research sweep

Related sources

• Prompt-to-host RCE follow-up

Tags

• ai-agent
• sandbox-escape
• rce
• configuration
• trust-boundary

This page is the archive's own catalogue record. The research is the work of Ilan Kalendarov, Ben Zamir and Elad Beber, first published at the original source. Preserved copies are kept so the citation survives its host.