Capability and Trust Matrix

Verdict, evidence trust, and decision capability are independent. Always assess all three.

Decision matrix

Goal	Entry point or evidence	Main prerequisites	Maximum supported claim in this slice	Does not prove
Author a business example	ScenarioVerifier.verify(...)	Scenario, fresh bootstrap, real graph and operators	AUTHORING	Source identity or CI-safe execution
Run Cases in JUnit	BlogeScenarioTests.fromDirectory(...)	Directory preflight and stable Case IDs	AUTHORING	Aggregate governance
Freeze one Suite to source	Source-bound Suite evidence	Clean Git worktree, tracked inputs, ignored output, external receipt	Usually AUTHORING	Team or release authority by itself
Prove Fixture fidelity	v4 Suite evidence	Declared effect boundary and complete control observations	Input to higher assessment	Undeclared ambient effects
Verify retained durable behavior	Durable v2–v4 evidence	Supported action plan, checkpoint/session evidence, source binding	Up to TEAM_GATE only with v4 prerequisites	CANCEL, streaming durable, production recovery certification
Aggregate one project	VerificationProjectRunner.runGovernedEvidence(...)	Same commit, governance selection, complete child inventory and receipts	Up to TEAM_GATE	Cross-reactor or external identity assurance
Aggregate direct Maven modules	VerificationReactorRunner.runGovernedEvidence(...)	Supported root coordinates and complete direct-module inventory	TEAM_GATE	Nested reactors, profile-derived coordinates, release qualification
Bind an independent reference model	Oracle v10/v11 evidence	Exact model closure, isolated child JVM, content-disjoint facts	Contributes to governed assessment	Semantic independence of copied logic
Add trusted external provenance	Ed25519 external attestation v1	Explicit trusted public key and exact governed receipt	GOVERNED_GATE	Signer correctness or undeclared behavior
Attribute one controlled change axis	DSL or BLOGE toolchain comparison v1	All other inputs frozen; BLOGE-owned authority for toolchain comparison	Still AUTHORING standalone	General regression freedom
Qualify a release decision	Release attestation v1 plus exact governed and comparison receipts	Governed evidence, attributable comparison, trusted signer, exact axis binding	RELEASE_QUALIFIED	Production correctness outside declared evidence

Cumulative gates

Claim	Required character of the evidence
AUTHORING	Useful for local contract development, but at least one higher-gate requirement is missing
TEAM_GATE	Source-bound, Policy-continuous, replay-assured, CI-safe evidence with complete declared effect boundaries
GOVERNED_GATE	Team-grade project evidence plus trusted external provenance and Oracle independence facts
RELEASE_QUALIFIED	Governed evidence plus an attributable comparison and a release attestation bound to the exact receipts and change axis

VerificationClaims.assess(...) calculates the result and lists missingEvidence. Callers cannot promote a claim by setting a field or by treating PASS as a capability.

Preview boundary

BLOGE Verify does not provide an OS, network, or external-service sandbox. Provider declarations and fingerprints are evidence about the declared boundary, not proof that no ambient dependency exists. A finite Property run is not a mathematical proof, and a passing comparison only covers the frozen inputs and observed business facts.