Skip to main content

DeepSeek-Reasonix EUVDEUVD-2026-88756

| CVE-2026-102437 HIGH
OS Command Injection (CWE-78)
2026-09-29 GitLab GHSA-8rx4-qgjw-3jvq
7.8
CVSS 3.1 · Vendor: GitLab
View

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain

  1. Access
    Craft repo with malicious .gitattributes + .git/config
  2. Delivery
    Deliver repo to victim (clone/share)
  3. Exploit
    Victim opens workspace-changes diff viewer
  4. Execution
    git invokes filter.clean/smudge command
  5. Impact
    Arbitrary command execution as local user

Vulnerability Assessment

Weakness
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization OS command sanitization, avoid shell exec
MITRE ATT&CK

2 techniques across 1 tactic: Execution

The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up.

Exploitation Conditions

Requires the attacker to control repository content - specifically a crafted .gitattributes registering a clean/smudge filter plus a .git/config binding that filter to an arbitrary command - and requires the victim to open that repository's workspace-changes diff viewer in the DeepSeek-Reasonix (Reasonix Studio) desktop app, which invokes git diff and triggers the filter. Limiting factors: exploitation is strictly local (AV:L), needs user interaction (opening/viewing the malicious repo), and depends on the victim trusting and loading an attacker-supplied repository; there is no remote or unauthenticated network path despite PR:N.

Impact

This is a genuine local privilege/code-execution risk but requires attacker-controlled repository content plus a victim action, so it is not a mass-exploitable network flaw. The CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H, base 7.8) is internally consistent: exploitation is local, requires no authentication to the app itself, but hinges on user interaction - the victim must open a malicious repository in the desktop app's workspace-changes diff viewer. The PR:N is technically correct (no app credentials needed) but is effectively gated by UI:R plus the prerequisite that the attacker controls the repo's .gitattributes and .git/config, which materially narrows real-world exploitability. No EPSS score, no CISA KEV listing, and no explicit POC flag are provided; a GitHub Security Advisory (GHSA-grg2-7gc6-36m6) exists but the input does not confirm exploit code, so exploitation status is 'no public exploit identified at time of analysis.' Net: high technical impact (full arbitrary command execution as the user) but a narrow, interaction- and configuration-dependent trigger against a niche product - a moderate real-world priority, not a widespread emergency.

Exploit ScenarioDerived

An attacker crafts a malicious Git repository containing a .gitattributes file that registers a clean/smudge filter and a .git/config that maps that filter to an attacker-supplied shell command, then lures a developer into cloning and opening it in DeepSeek-Reasonix (Reasonix Studio). When the victim views workspace changes in the diff viewer, the app invokes git diff, which triggers the filter and executes the embedded command with the user's privileges. No public exploit code is confirmed in the provided data, though a GitHub Security Advisory documents the issue.

This Week Schedule remediation this week. Why: HIGH severity (CVSS 7.0-8.9).
Patch available 1.0% exploitation probability (EPSS)

Get an e-mail when this CVE is exploited, patched or re-scored. Free account.

Alert me

AnalysisAIv1 – Sep 29, 2026

OS command injection in esengine DeepSeek-Reasonix (Reasonix Studio) allows a local attacker who controls a repository's contents to execute arbitrary commands with the privileges of the desktop application's diff viewer. The attacker supplies a crafted .gitattributes file that registers a clean/smudge filter plus a .git/config that binds that filter driver to an arbitrary shell command; when the victim opens that repository's workspace-changes diff view, the app's internal/gitcmd module runs git diff and the malicious filter is executed. Exploitation is strictly local (AV:L) and requires user interaction - the victim must open or view the attacker-supplied repository - and no public exploit code or confirmed active exploitation has been identified at time of analysis, so the risk is genuine but bounded rather than a mass-exploitable network flaw.

RemediationAI

Update DeepSeek-Reasonix to the release referenced in vendor advisory GHSA-grg2-7gc6-36m6 (https://github.com/esengine/DeepSeek-Reasonix/security/advisories/GHSA-grg2-7gc6-36m6) as soon as a fixed build is published; because the CPE data uses a wildcard version, verify the exact patched version in that advisory rather than assuming the latest build is fixed, and if the advisory lists only an upstream fix rather than a tagged release, treat the fix as unreleased upstream code. Until a patch is confirmed, treat every untrusted repository as executable content: do not open repositories supplied by third parties in the desktop app's diff viewer, and if they must be inspected, open them on a disposable sandboxed or containerized host or VM with no access to credentials or sensitive data. Additional compensating controls with trade-offs: pre-sanitize untrusted repositories by deleting or neutralizing .gitattributes and any filter.* entries in .git/config before loading them into the app (this breaks legitimate Git LFS and other filter-backed repositories), or run the diff viewer with Git attribute/filter processing disabled at the command level, for example by clearing the attribute source (GIT_ATTR_NOSYSTEM plus an empty core.attributesFile) or overriding filter configuration with empty -c filter.<driver>.clean=/smudge= values, which prevents the injected command from executing but removes filter-based normalization and may produce misleading diffs for repositories that rely on filters. Reducing the app's process privileges (running it as a low-privileged account, denying it access to SSH keys, tokens and cloud credentials) limits the blast radius because injected commands inherit the app's privileges. No vendor-released patch version has been confirmed from the available data, so monitor the advisory URL for the corrected release.

Affected VersionsENISA

DeepSeek-Reasonix 0 <2.21.0

Affected Packages

euvd deepseek_reasonix fix: 2.21.0

Source: OSV (Open Source Vulnerabilities)

Affects your stack

Affects 1 package in euvd

Checking this CVE against your own stacks and servers is part of the Basic plan and up.

Technical ContextAI · show details

The flaw is a classic OS command injection (CWE-78) rooted in Git's extensible filter mechanism. Git supports clean and smudge filters declared in .gitattributes (for example '*.txt filter=myfilter') and bound to shell commands in .git/config under filter.<driver>.clean / filter.<driver>.smudge or filter.<driver>.process. When git diff compares a worktree file against the index it normalizes the worktree content by running the configured clean filter, which means a diff operation - not just checkout or add - can spawn an attacker-chosen command through the shell. DeepSeek-Reasonix's internal/gitcmd component invokes git diff to render the workspace-changes view but does not neutralize or strip repository-supplied filter configuration, so command strings taken from attacker-controlled repository metadata reach a shell. Both inputs needed for exploitation (.gitattributes and .git/config) live inside the repository an attacker can distribute, and the CPE entry cpe:2.3:a:esengine:deepseek-reasonix:*:*:*:*:*:*:*:* uses a wildcard version, indicating no specific patched build has been enumerated in the CPE data. The reported-by source is GitLab and the vendor advisory is GHSA-grg2-7gc6-36m6 on the esengine/DeepSeek-Reasonix project.

Know when Deepseek Reasonix gets its next vulnerability

  • Watch products and get an email when something new is published
  • Check this CVE against your own stack and servers
  • Exploit and PoC references, threat intelligence from 22 sources

No credit card · 7-day full trial

Severity by source

Vendor (GitLab) PRIMARY 7.8 HIGH
AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
vuln.today AI 7.8 HIGH

Local trigger via a malicious repo opened by the victim (AV:L, UI:R); no app credentials needed (PR:N); git filter yields full arbitrary command execution as the user, so C/I/A:H.

3.1 AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
4.0 AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N

Primary rating from Vendor (GitLab).

CVSS VectorVendor: GitLab

Attack Vector
Local
Attack Complexity
Low
Privileges Required
None
User Interaction
Required
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
High

Lifecycle Timeline

2
Patch available
Sep 29, 2026 - 13:17 EUVD
CVE Published
Sep 29, 2026 - 12:06 cve.org
HIGH 7.8

DescriptionCVE.org

OS Command Injection in internal/gitcmd (git diff filter.clean/smudge invocation) in esengine DeepSeek-Reasonix (Reasonix Studio) allows a local attacker who controls repository content (.gitattributes + .git/config) to execute arbitrary commands via the desktop app's workspace-changes diff viewer.

Technical POC Analysis

Evidence base: GitHub advisory GHSA-grg2-7gc6-36m6 (esengine/DeepSeek-Reasonix).

Full technical POC analysis - root cause, step-by-step exploitation and patch breakdown - is part of the Pro and Business plans.

Executive ViewAI

Why It Matters

A high-severity command injection flaw in esengine DeepSeek-Reasonix (Reasonix Studio) lets a crafted repository execute arbitrary commands with the privileges of the desktop application's diff viewer. The risk is concentrated on developer workstations and source-code integrity, because a user must open the malicious repository and no vendor patch is currently available. Leadership should prioritize preventing users from opening untrusted repositories while tracking the vendor for remediation.

Business Impact: High
Compensating Controls

No vendor-released patch identified at time of analysis; interim mitigations include prohibiting opening untrusted repositories in Reasonix Studio, disabling or sandboxing its diff-viewer and git integration where possible, applying least privilege to the desktop application, and monitoring for unexpected child processes spawned during diff viewing.

Threat Intelligence Overview

EPSS Score
1.0%
Top 39% exploitable
Prevalence
Low
niche esengine developer/AI desktop app, minimal production footprint
Data Quality
5 sources
Confidence: Medium

How this decision was reached

The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.

SignalValue
Severity HIGH
CISA KEV No
Public POC No
EPSS 1.0%
Patch available Yes
Active threat No
Deciding rule
HIGH severity (CVSS 7.0-8.9)
What every action level means
Emergency Immediate action required. Criteria: CRITICAL + in CISA KEV (actively exploited) without patch, OR CRITICAL + EPSS > 30% without patch, OR CRITICAL + active threat signals without patch.
Act Now Prioritize within 24-48h. Criteria: in KEV with patch available, OR CRITICAL + public POC, OR HIGH + EPSS > 10%, OR CRITICAL + active threat, OR any CRITICAL severity.
This Week Schedule remediation this week. Criteria: HIGH severity (CVSS 7.0-8.9). Patch within normal change window.
This Month Plan within 30 days. Criteria: MEDIUM severity (CVSS 4.0-6.9), lower exploitation probability. Include in regular patch cycle.
Monitor Track and reassess. Criteria: LOW severity (CVSS 0.1-3.9), minimal exploitation risk. No immediate action needed.

SSVC AssessmentENISA

SSVC Decision
Scheduled
Exploitation
None
System Exposure· derived
Small
Automatable
No
Human Impact
High
Technical Impact
Total

Priority Score

40
Low Medium High Critical
KEV: 0
EPSS: +1
CVSS: +39
PoC: 0
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

OS command injection occurs when an application passes unsanitized user input directly into system shell commands.

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy