Skip to content

Security: MiviaLabs/mivia-agent

SECURITY.md

Security Policy

mivia is a local CLI agent. It reads, searches, and edits files in your project, and it runs the commands you allow. The security boundary is therefore critical: mivia must never read a secret it is not meant to see, must never expose a credential, and must never let an untrusted file change what it is allowed to do.

Reporting a vulnerability

Do not open a public issue for a security problem.

Report privately through the GitHub Security Advisory flow:

  1. Open the repository on GitHub.
  2. Select Security → Report a vulnerability.
  3. Describe the issue and include a minimal reproduction.

You can report any issue in scope below. We aim to acknowledge reports within 5 business days and to ship a fix or a mitigation within 30 days.

Scope

These areas are in scope for private reporting:

  • Secrets handling: API keys, env files, redaction, and log output.
  • Command execution: the run_command allowlist and blocklist, workspace confinement, and environment variable filtering.
  • File tools: path traversal, symlink handling, and workspace boundary escapes.
  • Lifecycle hooks: the trust model for PreToolUse, PostToolUse, and Stop scripts.
  • The provider boundary: what prompt and context data leaves the machine, and where it goes.

Anything outside this list is a normal bug: open a regular issue with the bug report template.

What we ask of you

  • Remove all secrets from your report before you submit it.
  • Include the output of mivia version --json and your operating system.
  • Do not publish the vulnerability until we ship a fix or decline to fix it.

Safe harbor

We will not pursue a claim against a researcher who reports in good faith, who does not access data beyond what the report needs, and who does not disrupt the service or harm other users.

There aren't any published security advisories