Skip to content

Security: blueraai/bluera-knowledge

SECURITY.md

Security

Reporting Vulnerabilities

If you discover a security vulnerability, please report it by opening a GitHub issue or contacting the maintainers directly.

Security Configuration

HTTP Server Binding

By default, the HTTP server binds to 127.0.0.1:3847 (localhost only). This means:

  • The server is only accessible from the local machine
  • External network access is blocked by default

To expose the server to the network (not recommended for untrusted environments):

bluera-knowledge serve --host 0.0.0.0 --port 3847

Warning: When binding to 0.0.0.0, the API is accessible to any device on the network. Only use this in trusted environments.

CORS Configuration

The HTTP API allows requests from any origin by default. This is appropriate for local development but should be restricted in production deployments.

Setup Command

The setup repos command clones repositories from a hardcoded list of trusted Anthropic/Claude repositories. The URLs are defined in src/defaults/repos.ts and are not user-configurable.

Security Practices

Input Validation

  • All MCP tool inputs are validated using Zod schemas
  • HTTP API endpoints validate request bodies before processing
  • Store IDs use branded types with regex validation (/^[a-zA-Z0-9_-]+$/)

Subprocess Execution

  • Git operations use spawn() with argument arrays (not shell strings)
  • Python bridge uses JSON-RPC over stdin/stdout
  • No shell command string interpolation

Path Handling

  • File store paths are normalized to absolute paths using path.resolve()
  • Directory existence is validated before creating file stores
  • Symlinks are resolved where security-relevant

External Communications

  • All external API calls (npm, PyPI, crates.io, Go proxy) use HTTPS
  • No API keys or credentials are stored in the codebase
  • Claude CLI integration uses the CLI's own authentication

Dependencies

Run regular security audits on dependencies:

npm audit
pip-audit  # For Python dependencies

There aren’t any published security advisories