Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach MCP tool interface with low privilege
-
ExploitSupply crafted videoPath/outputDir or plant maliciously-named file
-
ExecutionBreak out of child_process.exec command string via $()/backticks/quote
-
ImpactExecute arbitrary commands as MCP server 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 | Exploitation requires the attacker to control or influence the videoPath or outputDir arguments passed to the resize-video, extract-audio, or get-video-info tools of mcp-ffmpeg (≤1.0.5), or to cause those tools to operate on an existing file or directory whose name already contains $(), backticks, or a closing double quote. The payload must therefore live inside a path/name that the server forwards to child_process.exec, which is why the CVSS carries AT:P and AC:H rather than a free-form unauthenticated input - the condition is a crafted-but-already-present filesystem name reachable by the tool. Limiting factors: the attack vector is local (AV:L), PR:L indicates the attacker needs some level of access to reach the MCP tool interface, and the affected scope is restricted to mcp-ffmpeg versions up to and including 1.0.5; executed commands run only with the privileges of the MCP server user. |
| Impact | CVSS 4.0 rates this 7.3 (High), but the vector is telling: AV:L (local), AC:H, AT:P (a specific attack requirement must be present), and PR:L - this is not a trivially remote, unauthenticated RCE. The technical impact is high and well-supported by evidence (VC:H/VI:H/VA:H): successful injection executes arbitrary shell commands as the MCP server user, i.e. full compromise of the host account running the server. Against that, the exploitation conditions are constrained: an attacker must influence the videoPath/outputDir inputs to the resize-video, extract-audio or get-video-info tools, or cause the server to process an existing file/directory whose name already carries $(), backticks or a closing double quote - which is precisely what AT:P and AC:H encode here, grounded in the description's 'existing file or directory name' wording. No CISA KEV listing is present and no EPSS score was provided, so widespread-exploitation probability cannot be quantified from this data; absence of EPSS is a genuine gap, not a low score. The vuln.today decision label is 'this_week' (sourced from VulnCheck), consistent with a serious-but-not-emergency priority. Net: a high-impact flaw whose real-world urgency is tempered by a local attack vector, a required privilege level, and a non-trivial precondition - prioritize it where untrusted or attacker-influenced input can reach these MCP tools, but it does not carry the profile of a mass-exploitable internet-facing bug. |
| Exploit ScenarioDerived | An actor who can invoke the MCP server's resize-video, extract-audio or get-video-info tools supplies a videoPath or outputDir containing shell metacharacters such as $(id) or backticks, or causes the server to process a file/directory whose name already embeds them; when the server assembles the ffmpeg command and runs it through child_process.exec, the injected shell command executes as the MCP server user. No public exploit code was identified in the available data, though the CVE description and VulnCheck advisory describe the technique in concrete detail. Because the vector is local with a required precondition, the attacker must already be able to shape those path inputs or plant the maliciously named filesystem entry. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 28, 2026
OS command injection in the mcp-ffmpeg server (versions through 1.0.5) lets an attacker who can influence the videoPath or outputDir arguments of the resize-video, extract-audio, or get-video-info tools break out of the child_process.exec command string and run arbitrary shell commands as the user running the MCP server. The breakout characters ($(), backticks, or a closing double quote) must be carried inside a path or an already-existing file/directory name that the tool forwards into the exec string, which is why the CVSS 4.0 vector is scored 7.3 High with AV:L, AC:H, AT:P and PR:L rather than as a trivial remote unauthenticated RCE. No CISA KEV listing and no public exploit code were identified at time of analysis; the priority label is 'this week', and the practical blast radius is bounded by the privileges of the MCP server account.
RemediationAI
No vendor-released patch or fixed version is identified in the available data - the issue is tracked at https://github.com/bitscorp-mcp/mcp-ffmpeg/issues/8 and the VulnCheck advisory at https://www.vulncheck.com/advisories/mcp-ffmpeg-through-1.0.5-os-command-injection-via-videopath-and-outputdir, so confirm the current release status there before upgrading. Until a fixed release exists, treat the following as compensating controls rather than complete fixes, each with a trade-off: restrict who can reach the MCP tool interface and run the server under a dedicated low-privilege account, since injected commands inherit only the server user's rights (cost: reduced functionality for legitimate multi-user access); disable or gate the resize-video, extract-audio and get-video-info tools behind a trusted-input wrapper so that videoPath and outputDir only ever reference paths created or vetted by the operator (cost: those three tools become unusable for untrusted or agent-chosen filenames); and remove or rename any existing media files or directories whose names contain $(), backticks, or quote characters in directories the server is allowed to read or write (cost: manual filesystem hygiene, must be repeated as new files arrive). A durable fix requires changing the source to avoid shell interpolation - passing arguments as an array to child_process.execFile or spawn instead of building a command string for exec - which is a code change the operator must maintain if upstream does not ship it.
Affected VersionsENISA
Affected Packages
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
mcp-ffmpeg is a Model Context Protocol server that exposes ffmpeg operations as callable tools (resize-video, extract-audio, get-video-info) to an MCP client or agent. The flaw is CWE-78: the server composes shell command strings from caller-supplied videoPath and outputDir values and hands them to Node.js child_process.exec, which invokes a shell, so unescaped shell metacharacters in those strings are interpreted rather than treated as literal filename characters. The upstream reference points to src/mcp-ffmpeg.ts line 184 in commit 2ec04e4, and the public issue is bitscorp-mcp/mcp-ffmpeg#8. The CVSS 4.0 vector (AV:L/AC:H/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H, with no subsequent-system impact) and the analyst-assessed CVSS 3.1 equivalent (AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H) encode the real prerequisite: a crafted path must be presented to the tool, or the tools must be pointed at an existing file or directory whose name already contains $(), backticks, or a closing double quote, so the injected fragment lands in the shell command. The advisory does not describe a remote, anonymous HTTP entry point; reachability depends on how the MCP server is deployed and who can call its tools.
Know when Mcp Ffmpeg 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
CVSS:4.0/AV:L/AC:H/AT:P/PR:L/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
Local vector, requires crafting or a pre-existing path with shell metacharacters passed to exec (AC:H), low privilege to reach the MCP tools; full host command execution.
Primary rating from Vendor (VulnCheck).
CVSS VectorVendor: VulnCheck
DescriptionCVE.org
mcp-ffmpeg through 1.0.5 contains an OS command injection vulnerability that allows attackers to run shell commands by supplying crafted videoPath or outputDir values to the resize-video, extract-audio and get-video-info tools. Attackers can embed $(), backticks or a closing double quote in an existing file or directory name to break out of the child_process.exec command string and execute commands as the MCP server user.
Executive ViewAI
A high-severity OS command injection vulnerability in the mcp-ffmpeg server (versions up to 1.0.5) enables an attacker with local access and low privileges to execute arbitrary shell commands by manipulating path arguments in video processing tools. This compromises the MCP server account and potentially any data or systems accessible to it, though exploitation requires specific conditions and no active exploitation has been observed. Leadership should prioritize remediation to prevent potential unauthorized access and data exposure.
No vendor patch identified at time of analysis. Interim mitigations include: strict input validation to reject shell metacharacters in path arguments, running the MCP server under a least-privilege account, restricting local access to authorized users, and monitoring for anomalous command execution.
Threat Intelligence Overview
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.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
OS command injection occurs when an application passes unsanitized user input directly into system shell commands.
EUVD-2026-110846
GHSA-cqrv-cwxw-xcq4
NVD
CVE.org
EPSS
Exploit-DB