Skip to main content

argocd-mcp CVE-2026-108585

| EUVDEUVD-2026-96372 MEDIUM
Path Traversal (CWE-22)
2026-10-10 VulnCheck GHSA-f7cg-qc77-hh95
5.3
CVSS 4.0 · Vendor: VulnCheck
View

Attack ChainAIDerived

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

  1. Access
    Access MCP client or inject model prompt
  2. Delivery
    Call delete_application with ../repositories/ value
  3. Exploit
    Server concatenates path and sends authenticated DELETE
  4. Execution
    Argo CD deletes repositories, clusters, or projects
  5. Impact
    Loss of managed GitOps resources within token RBAC

Vulnerability Assessment

Weakness
CWE
OWASP Top 10:2025
A01:2025 - Broken Access Control
ASVS 5.0
V5 - File Handling Path canonicalization, traversal prevention
MITRE ATT&CK

2 techniques across 2 tactics: Discovery, Collection

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

Exploitation Conditions

Requires a running argocd-mcp server (version through 0.9.0) connected to an Argo CD instance with a valid API token loaded (PR:L), and the ability to invoke the delete_application tool - either as a direct MCP client or indirectly via prompt injection of an LLM connected to the server. The malicious trigger is an applicationName containing dot-segments (e.g. '../repositories/') that smuggles the request onto a different Argo CD API route. Limiting factors: no special non-default configuration is needed, but impact is strictly bounded by the token's RBAC permissions (a read-only or application-scoped token cannot delete repositories/clusters/projects), the MCP server must be deployed and token-authenticated, and there is no confidentiality exposure - only deletion of resources the token is already permitted to modify.

Impact

The input provides only a CVSS 4.0 vector (5.3, Medium) - CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:L/VA:L/SC:N/SI:N/SA:N; no CVSS 3.1 vector is supplied. Signals indicate network reach (AV:N) with low attack complexity (AC:L) but authenticated access (PR:L): the argocd-mcp server must already hold a valid Argo CD API token, so this is not an unauthenticated remote flaw. Impact is bounded - no confidentiality loss (VC:N) and only low integrity/availability impact (VI:L/VA:L) constrained to whatever the token's RBAC role permits. No EPSS score is present in the input and there is no CISA KEV listing, so there is no confirmed active exploitation. A VulnCheck advisory plus a public hackMD technical writeup and direct source-code references exist, meaning the technique is documented, but no packaged public exploit is flagged in the input. The most important real-world amplifier is the prompt-injection angle explicitly called out in the description: in agentic deployments an attacker who can influence the LLM's input - not just a direct MCP client - can trigger destructive DELETEs, which raises practical risk above what the raw 5.3 suggests for teams running this in automated pipelines. Net: a genuine but moderate, authentication- and RBAC-gated issue, elevated for agentic/LLM-exposed deployments and lower for tightly scoped tokens; EPSS data is missing and should be verified before final prioritization.

Exploit ScenarioDerived

An attacker with access to the MCP client, or who can inject a crafted prompt into an LLM wired to the argocd-mcp server, invokes the delete_application tool with applicationName set to a dot-segment value such as '../repositories/<repo>'. Because the value is concatenated into the API path without validation, the server issues an authenticated DELETE to an unintended Argo CD endpoint, destroying repositories, clusters, or projects within the token's RBAC scope. No packaged public exploit is identified in the input, though a VulnCheck advisory and a hackMD writeup describe the route-smuggling technique in detail.

This Month Plan within 30 days. Why: MEDIUM severity (CVSS 4.0-6.9).
No patch listed yet 0.3% exploitation probability (EPSS)

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

Alert me

AnalysisAIv1 – Oct 10, 2026

Path traversal in the argocd-mcp (Argo CD MCP Server) delete_application tool through 0.9.0 allows authenticated MCP clients to smuggle single DELETE requests onto unintended Argo CD API routes, so an applicationName value such as '../repositories/' can cause deletion of repositories, clusters, or projects instead of an application. The attacker must be a valid MCP client - or an LLM manipulated by prompt injection - operating with a token that already carries the corresponding RBAC permissions, so it is authenticated (PR:L) and impact is bounded to destructive writes (I:L/A:L) with no confidentiality exposure. No CISA KEV listing or public exploit code has been flagged in the available intelligence, though a public technical writeup and the implicated source lines are referenced; CVSS 4.0 rates this 5.3 (Medium) and the independent 3.1 assessment is AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:L.

RemediationAI

No vendor-released patched version is identified in the available data, so treat 0.9.0 and earlier as vulnerable and watch the upstream repository (https://github.com/argoproj-labs/mcp-for-argocd) and the VulnCheck advisory (https://www.vulncheck.com/advisories/argocd-mcp-through-0.9.0-path-traversal-via-delete-application-tool) for a fixed release; if you build from source, apply or backport the normalization fix in the delete_application path (src/argocd/client.ts lines 115-139 / src/argocd/http.ts lines 86-91) so applicationName is strict-validated against a name pattern rejecting '/' and '..' segments. Until then, the highest-value compensating control is least privilege: issue the MCP server an Argo CD token scoped only to the specific applications and verbs it needs, with delete rights on repositories, clusters, and projects withheld - this fully neutralizes the traversal because the smuggled route is rejected by RBAC, at the cost of blocking legitimate repository/cluster/project management through this integration. Additionally, keep the argocd-mcp server off untrusted networks and restrict which MCP clients and LLM agents may reach it (limiting prompt-injection reach, though it does not stop a determined direct client), and consider an API gateway or reverse proxy that logs and blocks DELETE requests to /api/v1/repositories, /api/v1/clusters, and /api/v1/projects from the MCP server's identity - a trade-off that can break legitimate automation using those endpoints. Treat MCP tool arguments as untrusted input in any wrapping layer, since a prompt-injected model can generate the malicious dot-segment value on the attacker's behalf.

Affected VersionsENISA

argocd-mcp 0 ≤0.9.0

Affected Packages

euvd argocd_mcp no fix

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

argocd-mcp is the Model Context Protocol server bridging LLM/MCP clients to an Argo CD instance's REST API, maintained in the argoproj-labs/mcp-for-argocd repository and tracked as CPE cpe:2.3:a:argoproj-labs:argocd-mcp. The root cause is CWE-22 (Improper Limitation of a Pathname to a Restricted Directory, i.e. path traversal / route smuggling): the delete_application tool interpolates the client-supplied applicationName directly into the outbound Argo CD API request path without normalizing or rejecting dot-segments. In the referenced code (src/argocd/client.ts around lines 115-139 and the request construction in src/argocd/http.ts around lines 86-91), the tool is expected to hit /api/v1/applications/{name}, but an input like '../repositories/' resolves to a sibling endpoint such as /api/v1/repositories, so the same authenticated DELETE verb is issued against a different resource class. The MCP server is designed to hold a single Argo CD API token, so the traversed request inherits that token's identity and any success/failure is dictated entirely by the token's RBAC policy.

Know when Argocd Mcp 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 (VulnCheck) PRIMARY 5.3 MEDIUM
CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:L/VA:L/SC:N/SI:N/SA:N
vuln.today AI 5.4 MEDIUM

Network-reachable tool call (AV:N/AC:L) but requires a valid loaded API token (PR:L); no confidentiality impact (C:N), with RBAC-bounded resource deletion driving low integrity and availability impact.

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

Primary rating from Vendor (VulnCheck).

CVSS VectorVendor: VulnCheck

Attack Vector
Network
Attack Complexity
Low
Attack Requirements
None
Privileges Required
Low
User Interaction
None
Vulnerable system: Confidentiality
None
Vulnerable system: Integrity
Low
Vulnerable system: Availability
Low
Subsequent system: Confidentiality
None
Subsequent system: Integrity
None
Subsequent system: Availability
None

DescriptionCVE.org

argocd-mcp (Argo CD MCP Server) through 0.9.0 contains a path traversal vulnerability in the delete_application tool that allows MCP clients to reach unintended API endpoints via unvalidated applicationName values. Attackers or prompt-injected models can supply dot-segment values like ../repositories/ to send authenticated DELETE requests deleting repositories, clusters, or projects within the token's RBAC permissions.

Threat Intelligence Overview

EPSS Score
0.3%
Top 80% exploitable
Prevalence
Low
niche, emerging MCP add-on for Argo CD; limited production footprint
Social Media
1
Sightings
Data Quality
6 sources
Confidence: High

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 MEDIUM
CISA KEV No
Public POC No
EPSS 0.3%
Patch available No
Active threat No
Deciding rule
MEDIUM severity (CVSS 4.0-6.9)
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· derived
Scheduled
Exploitation· derived
None
System Exposure· derived
Open
Automatable· derived
No
Human Impact· derived
Medium

Priority Score

26
Low Medium High Critical
KEV: 0
EPSS: 0
CVSS: +26
PoC: 0
Weaponized: 0
Threat intel: 0
Gov advisory: 0

Attack Technique

Path traversal exploits occur when applications use user-controlled input to construct file system paths without proper validation.

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