If you find a security issue in Domain Memory, please report it privately. Public issues for security bugs put users at risk before a fix is available.

Use GitHub's private vulnerability reporting:

1. Open https://github.com/mashware/domain-memory/security/advisories
2. Click "Report a vulnerability"
3. Fill in the form with as much detail as you can share.

This gives us a private discussion thread, a draft advisory, and an optional CVE if the issue warrants one.

If GitHub Security Advisories is not available to you, email mashware@gmail.com with:

• A clear description of the issue and its impact.
• Steps to reproduce, ideally with a minimal proof-of-concept.
• Your name and any preferred attribution (or a request to stay anonymous in the published advisory).

Please do not include exploitation details on a public issue, a public PR, or a public Discord/Slack/forum thread before the fix has shipped.

• We aim to acknowledge a report within 5 business days.
• We will keep you in the loop while we investigate and develop a fix.
• Once a patch is ready, we coordinate disclosure with you: published advisory, CVE if applicable, credit in the changelog and the advisory (unless you've asked to remain anonymous).
• Expected resolution timelines:
  • Critical: target fix in 7 days.
  • High: target fix in 30 days.
  • Medium / low: scheduled into a normal release.

Domain Memory is local-first software. The MCP server runs on the developer's machine, reads files in the project tree, and exposes an optional HTTP API on 127.0.0.1 (which can be put behind a bearer token via DOMAIN_MEMORY_HTTP_TOKEN).

In scope:

• The @mashware/domain-memory-server MCP and HTTP surfaces.
• The @mashware/domain-memory install flow (it writes files into the user's project — anything that escapes the project root is in scope).
• The @mashware/domain-memory-web viewer when bound to 0.0.0.0 or exposed via a tunnel.
• Path traversal, prompt-injection-as-persistence, dependency vulnerabilities surfaced by npm audit.

Out of scope (interesting but not security issues for this project):

• Issues that require the attacker to already have shell access on the developer's machine.
• Social-engineering scenarios where the developer is convinced to paste malicious content into their own knowledge store. The store is treated as trusted input — that is part of the threat model.
• Vulnerabilities in transitive dependencies that we cannot fix without an upstream release. We will track these but the upstream is the right place to file them.

Domain Memory is pre-1.0. Only the latest tagged version receives security fixes. If you need a patch on an older version, please open an issue and we'll discuss case by case.