Roadmap and FAQ
Roadmap
| Milestone | Scope | Status |
|---|---|---|
| M0 | Registry, CLI, risk indicator, scanner, evidence, audit chain, reports, API, agents, skills, adapters | Done (0.1.0) |
| M1 | Lockfiles, manifests with digests, pointers, diffs | Done (0.2.0); YAML and signed manifests planned |
| M2 | Organization policy packs, control and evidence lifecycle queues, exception expiry reminders | Planned |
| M3 | Accessible dashboard, dependency graph, review tasks | Planned |
| M4 | Database, SSO, RBAC, tenant isolation, encryption, signed audit | Planned |
| M5 | Read-only MCP server, CI adapters, model registry and data catalog connectors | Planned |
A roadmap item is not a feature. See ROADMAP.
FAQ
Does this make me compliant with the EU AI Act, ISO 42001 or NIST AI RMF? No. It organizes evidence and flags gaps in your own declarations. Compliance is a legal and organizational judgment.
Does it classify my system's risk category? No. The 0-100 score is an internal prioritization indicator. It is not an EU AI Act tier or any legal category.
Why is everything HUMAN_REVIEW_REQUIRED? By design. Tools and agents may prepare recommendations; named humans decide.
Why unknown instead of pass? Missing or stale evidence proves nothing. Showing it as unknown keeps gaps visible.
Does it send my data anywhere? No. There are no network calls. The optional API is loopback only.
Does it read my .env files or secrets? No. It notes that .env exists and refuses key-like files as evidence.
Is the audit report an independent audit? It is automated self-assurance: the tool runs hostile-input checks on its own code. A third-party audit needs a qualified human. The report says so.
Are the 20 agents running services? No. They are portable instruction files. Native execution in each IDE is not verified.
Why not YAML manifests? A YAML parser would be the first dependency. It is deferred until reviewed. JSON works today.
How do I prove the report was not edited? governsutra verify --report FILE recomputes the SHA-256 seal. This detects edits, not authorship. Signing is planned.
Can I use it in CI? Yes: npx governsutra-ai audit --out out returns exit code 2 on failure; evidence verify returns 2 on drift.
Which Node versions? 22 and newer; CI covers 22 and 24.