If you discover a security vulnerability in Leash, please report it by opening a GitHub Issue with the label security or email the maintainers directly.

Please do not publicly disclose the issue until it has been addressed.

• A clear description of the vulnerability
• Steps to reproduce (minimal .ae script if applicable)
• Potential impact (e.g., capability bypass, sandbox escape, privilege escalation)
• Suggested fix (if any)

Leash's security model is built on capability gating — programs start with zero permissions and must explicitly declare required capabilities. Security issues include:

• Capability bypass (gaining access to a capability not declared via package)
• Sandbox escape (executing arbitrary code outside the VM)
• Audit log evasion (making an IO/model call without being recorded)
• Determinism violations (@deterministic mode not enforced)
• Resource exhaustion (fuel/timeout limits bypassed)

Out of scope:

• Vulnerabilities in the host application that embeds Leash (those are the host's responsibility)
• Social engineering attacks

• Acknowledgment: within 48 hours
• Initial assessment: within 5 business days
• Fix (critical): within 7 days
• Fix (moderate/low): next release cycle

We maintain a list of security researchers who report valid vulnerabilities in CONTRIBUTING.md (with their consent).