If you discover a security vulnerability in NatShell, please report it responsibly:

Email: Open a GitHub Security Advisory (preferred) or email the maintainer directly via the contact on the GitHub profile.

Please include:

• Description of the vulnerability
• Steps to reproduce
• Potential impact

• Acknowledgment within 48 hours
• Fix timeline depends on severity — critical issues are prioritized for immediate patch release
• Credit given in the changelog unless you prefer anonymity

NatShell executes shell commands by design. The safety classifier (regex-based, in safety/classifier.py) is a guardrail, not a sandbox. Vulnerabilities in scope include:

• Bypasses of the blocked command classifier
• Path traversal in session/backup persistence
• SSRF bypasses in fetch_url
• Prompt injection that circumvents safety rules
• Information disclosure (API keys, credentials leaking)

Out of scope:

• Users intentionally running in danger safety mode
• Commands approved by the user via the confirmation dialog