CVE-2025-59458 is a high severity (CVSS 8.3) command injection in JetBrains Junie, where code execution was possible because of improper command validation. JetBrains lists fixed builds in the 243, 251 and 252 lines. Update to a build at or above the one for your line.
What's affected
The record says Junie builds before the following are affected: 252.284.66, 251.284.66, 243.284.66, 252.284.61, 251.284.61, 243.284.61, 252.284.50, 252.284.54, 251.284.54, 251.284.50, 243.284.54 and 243.284.50.
The list covers several release lines and variants. The JetBrains fixed issues page, referenced in the record, is the place to match your exact build to the right fix.
How to check if you're exposed
Open your JetBrains IDE and check the version of the Junie plugin under the plugins settings, then compare it with the build numbers above and the JetBrains page. If you manage many developer machines, ask your endpoint or IDE management tooling for installed plugin versions rather than relying on each developer to check.
If auto update is disabled in your environment, assume you are behind.
What to do now
- Update the Junie plugin to the fixed build for your IDE line.
- Make sure IDE and plugin auto update is on, or schedule updates for locked down machines.
- Review which commands the agent is allowed to run without asking, and tighten that list.
- Be careful opening unfamiliar repositories with an AI coding agent active.
- Keep secrets out of the environments where the agent runs commands.
Why it matters
The record maps this to CWE-77, command injection. The wording is that code execution was possible due to improper command validation, meaning the check that decides which commands an agent may run could be got around. The vector shows user interaction is required and the scope changes, with high impact on all three properties.
Coding agents combine two risky things: they read untrusted text, such as repository files and web pages, and they can run commands. The validation layer between those two is therefore a security boundary, and flaws in it are worth patching promptly. This CVE is not on the CISA KEV list.
Command injection occurs when user input is passed to an operating system shell or command line without being neutralised. The shell then treats part of the input as instructions. The durable fix is to avoid the shell, call library functions directly where possible, and pass arguments as separate values rather than one assembled string.
For defenders, the practical signals are unexpected child processes spawned by a web server, and outbound connections from application servers that have no business making them. Both are worth alerting on for any internet facing application, whether or not this particular flaw applies to you.
A dependency scan will not see this one, because it concerns a product or service rather than a package in your code. The practical step is an asset inventory that records which of these products you run, which version, and who owns it. You can look up the record for this CVE in the Safeguard Gold directory and use the free scan at gold.safeguard.sh/scan for the software you build yourself.
Quick check
Which Junie builds are fixed?
The record lists builds such as 252.284.66, 251.284.66 and 243.284.66 among several. Check the JetBrains fixed issues page for your line.
Why does user interaction appear in the vector?
The attack needs the user to do something, such as open content or accept an action, before the flaw is reached.
Sources
- JetBrains fixed security issues (listed in the record)
- NVD record data
Last reviewed: 18 October 2026.
