CVE-2026-77258: MCP Atlassian: path traversal exposes server files

Q: Is CVE-2026-77258 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-77258, increasing the risk of exploitation.

Q: How to fix CVE-2026-77258?

1) Upgrade mcp-atlassian to 0.22.0 or later (pip install -U mcp-atlassian). 2) Until patched, disable or block the Confluence upload_attachment tool (for example through tool allowlists or read-only mode). 3) Run the MCP server in a container or sandbox with a minimal filesystem, mounting no secrets, SSH keys or cloud credentials, as a non-root user. 4) Restrict who and which agents can call the server, and require human approval for file-handling tool calls. 5) Detection: audit Confluence for unexpected recent attachments (such as .env, id_rsa, config or passwd files), review MCP tool logs for upload_attachment calls with absolute or traversal paths, and rotate any secrets that were reachable by the server process if suspicious uploads are found.

Q: What systems are affected by CVE-2026-77258?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, MCP servers, tool-calling LLM agents, enterprise knowledge-base integrations.

Q: What is the CVSS score for CVE-2026-77258?

CVE-2026-77258 has a CVSS v3.1 base score of 7.7 (HIGH). The EPSS exploitation probability is 0.37%.

GHSA-93xw-j965-9mx3HIGHPoC AVAILABLECISA: TRACK*

Published September 22, 2026

CISO Take

MCP Atlassian versions before 0.22.0 let any authorized Confluence MCP caller pass an arbitrary server-local file path to the upload_attachment tool, which reads the file and uploads it to Confluence, disclosing anything the server process can read (CWE-22). The CVSS is 7.7 (High) with a changed scope, because the leaked data, such as .env files, SSH keys and API tokens, sits outside the vulnerable component. The exploitation likelihood score is low (EPSS 0.0034, around the 75th percentile), the flaw is not in CISA KEV, and SSVC says TRACK*, but a public PoC exists and exploitation needs only low privileges and no user interaction. Exposure is also reachable through an LLM agent: a prompt-injected agent can call the tool on an attacker's behalf. The package has only 1 downstream dependent, but 38 other CVEs have been recorded against it, which suggests weak security hygiene. Upgrade to mcp-atlassian 0.22.0 or later, and run the server under a least-privileged account with no access to secrets.

Sources:NVDGitHub AdvisoryEPSSATLAS

What is the risk?

High severity (CVSS 7.7) with network reach, low complexity, low privileges, no user interaction and a changed scope; confidentiality impact is high, with no integrity or availability impact. Real-world exploitation likelihood is moderate to low (EPSS 0.0034, not in KEV, no Nuclei template), but a public PoC lowers the barrier. Risk rises sharply when the MCP server runs with broad file permissions, in shared environments, or is exposed to untrusted agents or users. A prompt-injected agent could trigger the file read without a human attacker holding direct credentials. The package's history of 38 other CVEs raises the likelihood of further issues.

How does the attack unfold?

Initial Access

Attacker gains low-privileged access to the MCP Atlassian server or influences an agent that can call its tools.

AML.T0053

Exploitation

The attacker calls upload_attachment with a file_path pointing to a sensitive server-local file, since the path is not restricted to a workspace.

AML.T0037

Exfiltration

The server reads the file and attaches it to a Confluence page, where the attacker downloads it.

AML.T0086

Impact

Stolen secrets and credentials enable lateral movement into cloud, repository and Atlassian environments.

AML.T0055

Initial Access

Attacker gains low-privileged access to the MCP Atlassian server or influences an agent that can call its tools.

AML.T0053

Exploitation

The attacker calls upload_attachment with a file_path pointing to a sensitive server-local file, since the path is not restricted to a workspace.

AML.T0037

Exfiltration

The server reads the file and attaches it to a Confluence page, where the attacker downloads it.

AML.T0086

Impact

Stolen secrets and credentials enable lateral movement into cloud, repository and Atlassian environments.

AML.T0055

What systems are affected?

Package	Ecosystem	Vulnerable Range	Patched
MCP Atlassian	pip	< 0.22.0	`0.22.0`
2 dependents85% patched~4d to patchFull package profile →

Do you use MCP Atlassian? You're affected.

How severe is it?

CVSS 3.1

7.7 / 10

EPSS

0.4%

chance of exploitation in 30 days

Higher than 32% of all CVEs

Source: EPSS v3 — FIRST.org

Exploitation Status

Exploit Available

Exploitation: MEDIUM

Sophistication

Trivial

Exploitation Confidence

medium

○CISA SSVC: Public PoC

○Public PoC indexed (trickest/cve)

Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AVNetwork

ACLow

PRLow

UINone

SChanged

CHigh

INone

ANone

What should I do?

1 step

1) Upgrade mcp-atlassian to 0.22.0 or later (pip install -U mcp-atlassian). 2) Until patched, disable or block the Confluence upload_attachment tool (for example through tool allowlists or read-only mode). 3) Run the MCP server in a container or sandbox with a minimal filesystem, mounting no secrets, SSH keys or cloud credentials, as a non-root user. 4) Restrict who and which agents can call the server, and require human approval for file-handling tool calls. 5) Detection: audit Confluence for unexpected recent attachments (such as .env, id_rsa, config or passwd files), review MCP tool logs for upload_attachment calls with absolute or traversal paths, and rotate any secrets that were reachable by the server process if suspicious uploads are found.

What does CISA's SSVC say?

DecisionTrack*

Exploitationpoc

AutomatableNo

Technical Impactpartial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Data ExtractionData LeakageAgentPluginAML.T0037- Data from Local System AML.T0053- AI Agent Tool Invocation AML.T0055- Unsecured Credentials AML.T0086- Exfiltration via AI Agent Tool Invocation

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act

Art. 15-Accuracy, robustness and cybersecurity

ISO 42001

A.7.4-Quality of data for AI systems

NIST AI RMF

MANAGE 2.3-Responses to identified AI risks and unknown risks

OWASP LLM Top 10

LLM05-Supply Chain VulnerabilitiesLLM06-Excessive Agency

Frequently Asked Questions

What is CVE-2026-77258?

Is CVE-2026-77258 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-77258, increasing the risk of exploitation.

How to fix CVE-2026-77258?

1) Upgrade mcp-atlassian to 0.22.0 or later (pip install -U mcp-atlassian). 2) Until patched, disable or block the Confluence upload_attachment tool (for example through tool allowlists or read-only mode). 3) Run the MCP server in a container or sandbox with a minimal filesystem, mounting no secrets, SSH keys or cloud credentials, as a non-root user. 4) Restrict who and which agents can call the server, and require human approval for file-handling tool calls. 5) Detection: audit Confluence for unexpected recent attachments (such as .env, id_rsa, config or passwd files), review MCP tool logs for upload_attachment calls with absolute or traversal paths, and rotate any secrets that were reachable by the server process if suspicious uploads are found.

What systems are affected by CVE-2026-77258?

This vulnerability affects the following AI/ML architecture patterns: agent frameworks, MCP servers, tool-calling LLM agents, enterprise knowledge-base integrations.

What is the CVSS score for CVE-2026-77258?

CVE-2026-77258 has a CVSS v3.1 base score of 7.7 (HIGH). The EPSS exploitation probability is 0.37%.

What is the AI security impact?

Affected AI Architectures

agent frameworksMCP serverstool-calling LLM agentsenterprise knowledge-base integrations

MITRE ATLAS Techniques

AML.T0037Data from Local System

AML.T0053AI Agent Tool Invocation

AML.T0055Unsecured Credentials

AML.T0086Exfiltration via AI Agent Tool Invocation

Compliance Controls Affected

EU AI Act: Art. 15

ISO 42001: A.7.4

NIST AI RMF: MANAGE 2.3

OWASP LLM Top 10: LLM05, LLM06

What are the technical details?

Original Advisory

MCP Atlassian is a Model Context Protocol (MCP) server for Atlassian products (Confluence and Jira). Prior to 0.22.0, upload_attachment in src/mcp_atlassian/confluence/attachments.py accepts a caller-controlled file_path and opens the selected server-local file without restricting it to the workspace. A permitted Confluence MCP caller can upload the file as an attachment and disclose data readable by the server process. This issue is fixed in version 0.22.0.

Exploitation Scenario

An attacker with low-privileged access to an MCP client, or who plants an indirect prompt injection in a Confluence page or Jira ticket that a connected agent reads, causes the agent to call upload_attachment with file_path set to something like /home/svc/.env or ~/.ssh/id_rsa. The MCP server opens the local file and attaches it to a Confluence page the attacker can view. The attacker downloads the attachment, recovers API keys and cloud credentials, and uses them to pivot into source repositories, cloud accounts and other connected systems, then deletes the attachment to reduce traces.

Weaknesses (CWE)

CWE-22Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')Primary CWE-22Improper Limitation of a Pathname to a Restricted Directory

CWE-22 — Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal'): The product uses external input to construct a pathname that is intended to identify a file or directory that is located underneath a restricted parent directory, but the product does not properly neutralize special elements within the pathname that can cause the pathname to resolve to a location that is outside of the restricted directory.

[Implementation] Assume all input is malicious. Use an "accept known good" input validation strategy, i.e., use a list of acceptable inputs that strictly conform to specifications. Reject any input that does not strictly conform to specifications, or transform it into something that does. When performing input validation, consider all potentially relevant properties, including length, type of input, the full range of acceptable values, missing or extra inputs, syntax, consistency across related fields, and conformance to business rules. As an example of business rule logic, "boat" may be syntactically valid because it only contains alphanumeric characters, but it is not valid if the input is only expected to contain colors such as "red" or "blue." Do not rely exclusively on looking for malicious or malformed inputs. This is likely to miss at least one undesirable input, especially if the code's environment changes. This can give attackers enough room to bypass the intended validation. However, denylis
[Architecture and Design] For any security checks that are performed on the client side, ensure that these checks are duplicated on the server side, in order to avoid CWE-602. Attackers can bypass the client-side checks by modifying values after the checks have been performed, or by changing the client to remove the client-side checks entirely. Then, these modified values would be submitted to the server.

Source: MITRE CWE corpus.