CVE-2026-123851

Junie GitHub Action misses encoded prompt injection

JetBrains · JetBrains Junie GitHub Action

First documented disclosure · Review checkpoint

Summary

Encoded input can evade the agent's prompt-injection filter.

Severity and exploitation evidence

Medium · 5.4/10 · CVSS 3.1. Severity source.

Exploitation not established: The reviewed public sources describe the vulnerability and its conditions. Malicious exploitation has not been established in the reviewed sources.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The action did not normalize encoded input consistently before applying its prompt-injection protection.

Conditions required

The CNA describes network access and user interaction. JetBrains does not identify the affected workflow event.

Potential impact

Untrusted instructions may influence the coding agent. The CNA rates confidentiality and integrity impact as low.

Affected versions and configurations

Before 1.6.4.

Fix and mitigation status

1.6.4.

Practical next steps

Update the Junie action to 1.6.4 and review the permissions its workflows grant.

Operating lesson

Normalize equivalent input representations before applying safety checks.

Evidence and limits

JetBrains identifies internal issue JUNIE-3424 and the fixed action version in its public security table.

Sources