Export
The link always opens the latest version of this page.
Ask an AI about it
Open in ChatGPTA new chat that reads this entry from the site Open in ClaudeThe same, in ClaudeShare the link instead of a file, and they will always see the latest version.
Files are a snapshot of today. The live page has anything newer.
GitSpawn
CLI AI coding agents execute untrusted git configuration commands without user approval when processing repositories
- Weakness
- core.fsmonitor 1 source
GitSpawn is a vulnerability affecting cli AI coding agents that execute untrusted git configuration commands without user approval when processing repositories.
Affected Products
The vulnerability impacts multiple AI coding agents: Claude Code 1, Goose 1, Grok 1, Hermes 1, and Qwen Code 1.
Technical Details
The vulnerability exploits the core.fsmonitor git configuration option 1, which allows arbitrary command execution through git's filesystem monitoring mechanism. When these agents clone or interact with repositories containing malicious git configurations, they execute the configured commands without explicit user approval.
Attack Vector
An attacker can craft a repository with a malicious .git/config file that specifies a core.fsmonitor command. When an AI coding agent processes this repository—typically during code analysis, cloning, or repository inspection—git automatically executes the configured command in the agent's execution context, potentially leading to unauthorized code execution.
Impact
This vulnerability enables remote code execution on systems running affected AI coding agents, as the agents operate with the permissions of their execution environment and process untrusted repositories without sandboxing git configuration directives.
Facts 1
| Attribute | Value | Confidence | Sources | Cited |
|---|---|---|---|---|
| Weakness | core.fsmonitor | low | 1 | 1 |
Affected product 5
Sources 1
- GitSpawn: Untrusted repos can execute code via AI coding agents · Hacker News (YC)