HALCYON is a reference implementation intended to demonstrate and prototype a partitioned-trust architecture. It has not undergone an independent security audit. Do not deploy it as-is to protect production systems without your own review and threat modeling. See SPEC.md §9 for known residual risks and §0 for the threat model it is scoped to.

Please report suspected vulnerabilities privately rather than opening a public issue. Email: ideatrino@proton.me (replace before publishing).

Include, where possible:

• a description of the issue and the layer/module affected,
• steps or a minimal script to reproduce,
• the impact you believe it has.

You can expect an acknowledgement within a few days. Coordinated disclosure is appreciated; we will credit reporters who wish to be named.

In scope: the control layer (Layer A / taint + checker + interpreter), the certified screen (Layer C), the DP accountant (Layer B), the integrity primitives (Layer D), and the ensemble (Layer E). Out of scope: the correctness/behavior of any third-party model you plug into the backends, and the security of the machine running the code.