Portal Community
Open decision — single-use AppRole secret_id: if a tenant's AppRole is configured with secret_id_num_uses = 1 (HashiCorp's own recommended hardening pattern), this node's re-authenticate-and-cache design works until the cached client token expires, then fails every subsequent execution with HASHICORP_AUTH_FAILED — there's no secret_id left to log in with again. Two documented options exist, and an explicit choice is needed before shipping to any tenant using single-use AppRole credentials: See Authentication for how the current cache behaves.

Deliberately Not Implemented

AreaWhy it's out of scope today
Response caching for Secrets reads Caching a secrets-manager response is a security anti-pattern, not an oversight — this is a permanent design choice, not a gap to close later.
token/create Minting arbitrary new tokens with elevated policies is an admin operation, not something a workflow-automation credential should do. Not planned.
Kubernetes / LDAP / Okta / JWT / cert auth methods Only Token and AppRole are supported — the two realistic automation-credential shapes. No other auth backend is wired up.

Phase 2 Candidates

AreaStatus
Transit secrets engine (encrypt/decrypt as a service) Flagged as a Phase 2 candidate in the design docs, not committed scope. No third-party corroboration of demand was found for it at design time.

Before This Is Production-Ready

No live integration testing yet. Everything currently shipped is verified by compiling and by direct comparison against HashiCorp's own official Vault API documentation — not by exercising a running Vault server. Live integration testing against a real Vault instance (OSS, Enterprise, and HCP Vault Dedicated) is required before calling this node production-ready.

What Won't Change

The 4-resource/18-operation shape (secrets/token/lease/system) reflects a deliberate scope decision, cross-checked against HashiCorp's own API docs across three review passes covering standby-vs-sealed status codes, LIST-on-empty-path behavior, AppRole single-use secret_id semantics, and cas: 0 semantics. New operations, if added, are expected to follow the same partial-class-per-operation / one-service-per-resource pattern already in place.