Skip to content

Security: Clueless-Creations/riverpod-knowledge-graph

Security

SECURITY.md

Security Policy

Supported security model

This repository is designed for local, stdio-based MCP usage on a trusted workstation.

Supported today:

  • local docs-only MCP through riverpod-knowledge-graph mcp
  • local audit-enabled MCP through riverpod-knowledge-graph mcp-audit
  • local CLI usage for audit-repo and build-snapshot

Not supported today:

  • direct internet-facing MCP deployment
  • shared or multi-tenant remote hosting
  • Cloudflare Workers deployment without a dedicated transport and auth layer

Secure defaults

  • mcp is docs-only and does not expose audit tools
  • public MCP read tools are pinned to the configured corpus root
  • mcp-audit is explicit opt-in because it can inspect local repo files
  • bootstrap-codex only writes a local Codex config entry; it does not expose a remote service

Trust boundary

  • Snapshot read tools should only read the bundled corpus root configured at startup.
  • Audit tools are local filesystem readers and should only be enabled for trusted local repos.
  • Remote hosting would require authentication, authorization, request isolation, and a non-stdio transport before it should be considered safe.

Repo hygiene

This repo should never contain:

  • secrets, tokens, credentials, or private keys
  • machine-local config files with user-specific absolute paths
  • private customer data or repo snapshots that are not redistributable

Reporting

If you find a security issue in the package defaults or MCP surface, please open a GitHub security advisory or contact the maintainers through the repository’s security reporting flow if it is enabled. If a private channel is not available yet, avoid posting exploit details in a public issue.

There aren't any published security advisories