Executive Summary
Researchers disclosed an Amazon Kiro prompt-injection flaw where attacker-controlled repository content could make the agent send sensitive local information to an external endpoint through Kiro Powers.
What Happened
Security researchers disclosed a vulnerability in Amazon Kiro, an agentic integrated development environment, in which attacker-controlled repository content could influence the agent. The reported attack path could cause sensitive local information to be transmitted to an external endpoint through prompt injection and Kiro Powers.
Affected Products
- Amazon Kiro
Vulnerability / Attack Technique
Indirect prompt injection leading to data exfiltration
Technical Mechanism
Repository content carried instructions the Kiro agent followed, using Kiro Powers to transmit local data externally.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
The same Kiro issue reported by multiple outlets. RealGround analysis: agent extensions that can send data out need approval prompts when triggered by untrusted content.
Recommended Controls
- Update Kiro
- Require approval before agent extensions send data externally
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
The Hacker News: https://thehackernews.com/search/label/Application%20Security
Summary and analysis written by RealGround from the source above; last updated 2026-10-17.