xllama moves fast and only the latest tagged release receives security fixes.

Use GitHub's private vulnerability reporting: https://github.com/gianlucamazza/xllama/security/advisories/new

Please do not open public issues for suspected vulnerabilities.

Include: affected version/commit, build target (UWP/x64 or Linux), reproduction steps or PoC, and expected vs actual behaviour.

In scope:

• The LAN HTTP endpoint (SessionHub surface): request parsing, auth-less exposure defaults, path traversal via persisted state files
• MSIX packaging / AppContainer sandbox escapes reachable from app content
• Model-loading paths (GGUF / ONNX) parsing untrusted model files

Out of scope:

• Anything requiring retail-console modification or hardware attacks (the project ships Dev Mode only)
• Attacks that require an already-compromised dev machine
• Reports about bundled upstream code where the fix belongs upstream (llama.cpp, ONNX Runtime) — still welcome, but may be routed there

The LAN endpoint is opt-in and default OFF, binds to the local network only, and requires the app in the foreground (UWP PLM). It carries no authentication by design — it is a single-operator research tool, not a multi-user service. Reports demonstrating impact beyond "anyone on the LAN can call a documented endpoint" are especially valuable.

• Acknowledgement: within ~7 days
• Triage verdict: within ~30 days
• Fix or mitigation: best effort alongside regular releases