Please do not report security vulnerabilities through public GitHub issues, discussions, or pull requests.

Report privately through GitHub's built-in flow:

1. Go to the repository's Security tab → Report a vulnerability (Privately reporting a security vulnerability).
2. Describe the issue, the impact, and steps to reproduce.

This opens a private security advisory visible only to you and the maintainer.

Please include, where you can:

• The type of issue and the component/file involved.
• Steps to reproduce, or a proof of concept.
• The impact — what an attacker could do.
• Any suggested remediation.

XE Local AI Engine is a local-first desktop application: it runs a local web server on loopback, supervises local inference runtimes, and stores data in a per-user directory. Areas of particular interest:

• The local admin API surface (/api/local/v1) and its loopback/Host/Origin guarding.
• Handling of secrets at rest (node key, Data Protection key ring, encrypted columns/blobs).
• The agent process/container sandbox and Development Mode's code-execution boundary.
• Model/skill import and any path-traversal or untrusted-content handling.

This is an early-stage project maintained by one person in their spare time, so responses are best-effort rather than bound to a formal SLA. You will get an acknowledgement, and coordinated disclosure once a fix (or a documented mitigation) is available. Please give a reasonable window before any public disclosure.

• Release binaries are currently unsigned because the project does not yet have a signing certificate. Windows can show browser reputation and SmartScreen Unknown publisher warnings; Linux desktop environments or security tools can require an explicit trust/execute action for the AppImage. Certificate signing is planned.
• Official releases publish CHECKSUMS.sha256, RELEASE-MANIFEST.json, and RELEASE.spdx.json. Verify the platform artifact's SHA-256 before running it. Checksums and SPDX inventory improve integrity/transparency but do not provide publisher authentication equivalent to a trusted code signature.
• Development Mode executes repository code on your machine by design; treat it as running the target repo's own build/test commands. See the user guide's privacy/security notes.