Preliminary research

Attacks via OpenClaw: when your LLM can make RCE

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.

Shows an OpenClaw command-injection path that first makes normal web retrieval fail, encouraging the model to fall back to a shell-built curl command. Attacker-controlled redirect or page values then cross into the shell command line, turning indirect prompt influence over an ordinary web request into code execution on the agent host.

Record

Published by

Purpleshift

Date

21 April 2026

In the archive

• Collected in the 2026 preliminary research sweep

Tags

• ai-agent
• prompt-injection
• command-injection
• rce
• url-parsing
• owasp-a03-2021

This page is the archive's own catalogue record. The research is the work of Purpleshift, first published at the original source. Preserved copies are kept so the citation survives its host.