GHSA-vr7g-88fq-vhq3: Paperclip: OS Command Injection via Execution Workspace cleanupCommand
criticalvulnerability
security
Source: GitHub Advisory DatabaseApril 16, 2026
Summary
Paperclip AI v2026.403.0 contains a critical OS command injection vulnerability in its execution workspace lifecycle. An attacker can inject arbitrary shell commands into the `cleanupCommand` field (a parameter that specifies what to clean up after a workspace closes) via the `PATCH /api/execution-workspaces/:id` endpoint, and these commands are executed without any validation when the workspace is archived. In the default `local_trusted` deployment mode, this attack requires no authentication and works on Linux, macOS, and Windows systems with Git installed.
Classification
Attack SophisticationModerate
Affected Packages
@paperclipai/server@< 2026.416.0 (fixed: 2026.416.0)
Monthly digest — independent AI security research
Original source: https://github.com/advisories/GHSA-vr7g-88fq-vhq3
First tracked: April 16, 2026 at 08:00 PM
Classified by LLM (prompt v3) · confidence: 95%