Amazon Kiro IDE Data Exfiltration via Filename Prompt Injection and Kiro Powers Registry Fetching

Affected Vendor(s)

Affected Product(s)

Summary

The Amazon Kiro IDE is vulnerable to a data exfiltration issue that can be exploited through a prompt injection and abuse of the Kiro Powers features. By crafting a repository containing a directory with prompt injection instructions in its name, an attacker can coerce the application to visit an attacker controlled website while submitting sensitive local file contents.

Timeline

Discovered on
December 11, 2025
Disclosed to Vendor on
December 11, 2025
Published on
January 15, 2026

Credit

Blog Post

References

Learn how Mindgard can help you navigate AI Security

Take the first step towards securing your AI. Book a demo now and we'll reach out to you.