Cyata found path-validation and argument-injection flaws in Anthropic's mcp-server-git (CVE-2025-68143, -68144, -68145) that, chained with the Filesystem MCP server, allowed file access and code execution by anyone able to influence what the assistant reads. Anthropic removed the git_init tool and added path validation in fixed releases.
Why it matters
Even the protocol author's reference servers carried injection-reachable code execution paths.
Key facts
As stated in the sources, with where to find them.
- CVE-2025-68143 fixed in mcp-server-git version 2025.9.25; CVE-2025-68144 and CVE-2025-68145 fixed in version 2025.12.18, per The Hacker News.The Hacker News, CVE list
Findings that cite this record
No tracked finding cites this record yet.
Key questions this bears on
- Can prompt injection against AI agents be reliably defended?Not reliably. Bounding what untrusted input can trigger helps, but trust boundaries must survive delegation and tool execution.
- Where are deployed AI agents actually being exploited?Exploitation often reaches deployed agents through connectors, credentials, tools and packages.
Sources
Related records
Oct 29, 2026
Apr 15, 2026
May 7, 2026
Aug 1, 2025
Feb 5, 2026
Oct 13, 2026