The Java testing library jqwik released version 1.10.0 containing a hidden prompt injection command designed to manipulate AI coding agents by instructing them to delete project source code.

Key Points

• Version 1.10.0 of the jqwik testing library includes a method named printMessageForCodingAgents that outputs a malicious command to standard output.
• The command uses ANSI escape sequences to hide the text from human terminal users while remaining visible to automated CI logs and AI coding tools.
• The maintainer explicitly included this code as a form of "openly communicated resistance" against the use of generative AI in software development.
• Security researchers note that traditional supply-chain scanners fail to detect this behavior because the library performs no unauthorized system calls or file modifications.
• The project documentation now includes a warning that using jqwik with AI coding agents is strongly discouraged.

Why it Matters

This incident highlights a new class of supply-chain risk where legitimate dependencies are used to deliver prompt injections directly into the context windows of AI coding assistants. Because the malicious instructions are delivered through standard output rather than traditional code execution, they bypass existing security tooling and automated provenance checks.