Skip to content

Security: hexide-io/HexIDE

Security

SECURITY.md

Security Policy

HexIDE is pre-alpha software. We take security seriously and appreciate responsible disclosure.

Reporting a vulnerability

Please do not open a public issue for security problems.

Report privately via GitHub's private vulnerability reporting (Security → Advisories → "Report a vulnerability"). If private reporting is unavailable, email security@hexide.io with details and, if possible, a proof of concept.

We aim to acknowledge reports within a few days. As an unfunded community project, fixes are best-effort, but security issues are prioritised over feature work.

Supported versions

Only the latest default branch is supported. There are no released binaries yet.

Security-relevant surfaces

A few parts of HexIDE are worth calling out for reviewers:

  • Add-in system. HexIDE can load add-ins, which are .NET assemblies that run in-process with full trust. Add-ins are verified against a signing trust chain (see TRUST.md) and first load requires explicit user consent. Only install add-ins from publishers you trust. The bundled development trust chain is not the release chain and is rotated to an offline root before any signed release.
  • MCP automation server. A local HTTP server that can drive the IDE exists only in Debug builds (#if DEBUG), is off by default, and must be explicitly enabled with --server-port. It is compiled out of Release builds entirely — a distributed binary opens no port. It currently has no authentication, so never enable it on an untrusted network.
  • LSP server. A separate subprocess communicating over stdio that parses VB6/VBA source. It does not open a network socket.

Data handling

The core IDE does not phone home, collect telemetry, or transmit your projects anywhere — it reads and writes local files only. Optional add-ins (for example, an AI chat assistant) may contact external services that you configure; consult each add-in's documentation.

There aren't any published security advisories