mind-node is research-grade software. The Solid data plane (LDP, WAC/ACP, RDF, Solid-OIDC, DID auth) is conformance-tested against the official Solid Conformance Test Harness. The economic layers — the MIND token ledger, cross-server settlement, the hand-rolled weighted-BFT FEDSET consensus, and content-pinned contracts — have not undergone an external security or cryptography review, which the design documents themselves name as a prerequisite for production.

Do not run this with real funds or sensitive production data.MIND is a closed-loop unit and is not redeemable for fiat. Treat all token-, settlement-, and contract-related functionality as a prototype.

Pre-1.0. Only the latest main is supported. There are no security backports.

Please report suspected vulnerabilities privately — do not open a public issue for anything exploitable.

• Preferred: open a GitHub private security advisory ("Report a vulnerability") on the repository.
• Include: affected component/crate, version or commit, reproduction steps, and impact.

We aim to acknowledge reports within a reasonable timeframe. Because this is an experimental project maintained on a best-effort basis, we cannot commit to a fixed remediation SLA, but we will work with you on coordinated disclosure.

• Most subsystems are off by default behind explicit flags (--ledger, --ledger-transfers, --ledger-settlement, --fedset, --contract, --llm-*, …). With a default-flag server the attack surface is the Solid data plane plus the operator/admin control plane.
• Secrets are supplied via environment only (never CLI flags); see .env.example and deploy/solidrs.env.example.
• The operator/admin plane (/.admin) is control-plane only and never exposes or mutates pod contents — WAC/ACP still gate the data plane. Reports that demonstrate an operator reading pod contents, or any authz bypass, are especially valuable.