The module maturity bar

Every one of the 110 modules carries a maturity in trustweave-capabilities.json. This page is the rule those maturities are assigned by, so that a reader can disagree with a particular classification on the evidence rather than on taste.

Before this, 100 modules carried no maturity at all. A consumer reading the catalog could not tell “we assessed this and it is experimental” from “nobody has looked at it yet” — and the second is the one that matters, because it is the one where the absence of a warning means nothing.

The three ratings

stub — delivers no callable capability

Either every public entry point throws, or the module has no SDK surface at all (a build aggregator, an application shell). Depending on one for function will fail at runtime, by design: these fail closed rather than pretending.

Eleven modules: anchors:plugins:starknet, did:plugins:threebox, did:plugins:tezos, did:plugins:btcr, credentials:plugins:platforms:salesforce, credentials:plugins:platforms:servicenow, distribution:all, distribution:bom, and the three reference-wallet:android modules.

distribution:examples is deliberately not on this list: it carries about 2,500 lines of runnable example code with its own tests, so it is experimental like any other implemented module.

experimental — implemented and exercised, not qualified

There is a real implementation and tests run against it, but at least one requirement of supported is unmet. ModuleCapabilities.requireDeployment refuses these unless the host passes allowExperimental, so a production policy excludes them until someone says otherwise.

Ninety-nine modules — currently everything that is not a stub.

supported — qualified for production

All four of the following, and each has to be a thing someone can open and read:

  Requirement How it is evidenced
1 Coverage floor — ≥70% line and ≥60% branch for the module The merged Kover report
2 Security review — named in a dated review, findings resolved A file under docs/reviews/
3 Interoperability evidence — a retained artifact from a cross-stack or vector-based test A CI artifact from a named run
4 Documented operational limits — what it does not do, how it fails, where it stops scaling A runbook section under docs/operations/

Nothing is supported yet, and that is the honest answer

Thirty-two of the 110 modules meet requirement 1. None meets all four, so none is classified supported.

The gap is not coverage. It is that requirements 2–4 are evidence about a module that mostly does not exist yet: one cross-stack interoperability artifact is retained in CI (verifiable-intent), and no module has a documented operational-limits section of its own.

Classifying a module supported without that evidence would be the same error this repository has been correcting elsewhere — asserting a control that was never established. So the catalog says experimental, and this section says exactly what would change it.

What the GA core needs

These eight are the intended first supported set. Coverage figures are from the merged report.

Module Line coverage Still needs
kms:kms-core 82.9% ✅ security review · interop evidence · documented limits
credentials:plugins:verifiable-intent 90.0% ✅ security review · documented limits (interop evidence ✅)
did:plugins:key 68.2% coverage (+1.8) · security review · interop evidence · documented limits
did:plugins:web 21.5% coverage (+48.5) · security review · interop evidence · documented limits
did:did-core 58.4% coverage (+11.6) · security review · interop evidence · documented limits
credentials:credential-api 68.0% coverage (+2.0) · security review · interop evidence · documented limits
wallet:wallet-core 30.4% coverage (+39.6) · security review · interop evidence · documented limits
one KMS provider kms:plugins:aws 26.6% coverage (+43.4) · a completed custody qualification run · documented limits

The KMS provider row is the one that does not compress: a qualified custody profile needs the live qualification to have actually run, which needs an AWS account with the three keys and the IAM deny in place. That is the same blocker as A3 in the 13 September review.

Changing a classification

scripts/check-capability-coverage.py enforces two things: every module carries a maturity from {stub, experimental, supported}, and the count of unassessed modules can only fall. It is now zero, so the ratchet is closed — a new module must be classified in the commit that adds it.

The classification is reproducible rather than remembered: it is derived from per-module coverage and test evidence by the rule above. To promote a module, add the missing evidence, link it here, and change the entry — in that order.


This site uses Just the Docs, a documentation theme for Jekyll.