Arbitrary Command Execution in Goose via Unsafe Recipe Extensions

Executive Summary

This vulnerability in goose 1.37.0 allows arbitrary command execution through recipe stdio extensions and retry.checks without proper security checks. Attackers can create malicious recipes that run shell commands as the user executing goose, bypassing the recipe security scan which only checks for hidden Unicode tags in limited fields.

Useful 0 Not Useful 0

Detection Guidance

To detect this vulnerability, inspect goose recipe files for malicious extensions or retry checks. Check for `cmd` and `args` in stdio extensions and `retry.checks.command` fields. Review all recipe sources including files, deeplinks, and GitHub repos for these fields.

Useful 0 Not Useful 0

Impact Analysis

An attacker could craft a malicious recipe that executes commands on your machine without your consent when you run the recipe. This could allow writing files, installing malware, or performing other actions as your user account. The vulnerability affects all recipe distribution methods including files, deeplinks, and GitHub repositories.

Useful 0 Not Useful 0

Compliance Impact

This vulnerability could lead to unauthorized command execution, potentially exposing sensitive data. For GDPR, it may violate principles of data protection by design and default (Article 25) if personal data is accessed or exfiltrated. Under HIPAA, it could compromise protected health information if the goose tool processes such data, violating the Security Rule's integrity and access controls.

Useful 0 Not Useful 0

Mitigation Strategies

Immediately update goose to a patched version beyond 1.49.0. Disable untrusted recipe sources and avoid running recipes from unknown origins. Implement user consent prompts for extensions and retry checks. Expand security scans to cover all recipe fields.

Useful 0 Not Useful 0