AIKIDO-2026-835820
strands-agents-tools is vulnerable to Improper Neutralization of Input Used for LLM Prompting
88
High Risk
This Affects:
strands-agents-toolsTL;DR
The shell tool enforces a human consent gate before running OS commands, but exposes the non_interactive parameter as an LLM-controllable input. Through indirect prompt injection in untrusted web pages, messages, or files, the model can be influenced to set non_interactive to true and skip operator approval. This lets arbitrary OS commands run with the agent process's privileges on the host. The fix removes model control over the non-interactive behavior so the consent gate cannot be bypassed.
Who does this affect?
You are affected if you are using a version that falls within the vulnerable range and you register the shell tool on an agent that processes untrusted content.
Background info
strands-agents-tools is vulnerable to Improper Neutralization of Input Used for LLM Prompting in versions 0.0.1 - 0.7.6.
How to fix this
Upgrade the strands-agents-tools library to the patch version.
Links
GitHub Release
Are You Affected?
Connect your repositories to instantly see whether vulnerable or malicious packages exist in your codebase.
Free. No credit card required.
I consent to receiving marketing communications based on Aikido’s Privacy Policy.
SOC 2Compliant
ISO 27001CompliantISO 42001Compliant
FedRAMPImplementing© All Intel data is openly available and commercially licensed.