Executive brief
dvladimirov MCP is a Model Control Plane server used to integrate AI services with Git repositories and local filesystems. A security flaw in its Git search functionality allows a remote attacker to execute arbitrary commands on the host server. This could lead to a total compromise of the system, including unauthorized access to sensitive data, service disruption, or the modification of files.
Technical details
A command injection vulnerability exists in the `GitSearchRequest` function within `mcp_server.py` of the dvladimirov MCP project (up to version 0.1.0). The root cause is the improper neutralization of the `repo_url` and `pattern` arguments, which are interpolated into a shell command string using Python f-strings and executed via `subprocess.check_output` with `shell=True`. Specifically, the `search_repository` method in `mcp/git_service.py` passes the raw user-supplied pattern to a `grep` command. A remote, unauthenticated attacker can exploit this by sending a specially crafted POST request to the `/v1/models/git-analyzer/search` endpoint containing shell metacharacters (e.g., `;` or `"`). This allows for arbitrary code execution on the host operating system. As of the advisory date, no official patch has been released.
Affected products
- dvladimirov MCP 0.1.0
Timeline
- 2026-04-10: disclosed: Vulnerability reported to the vendor via GitHub issue
- 2026-04-28: advisory: NVD/VulDB advisory published