A green parameterized matrix without suite revision, dataset version, parameter bindings, build id, and run artifacts is incomplete evidence. Bind early, freeze inputs, package the full chain.

Data-driven testing for release evidence means versioned datasets are first-class inputs. Parameterized greens without a dataset version are incomplete evidence—not full traceability.

Before generating MCP Rigor acceptance tests, confirm approved tool, resource, or prompt contracts with inputs and expected behavior. Incomplete contracts yield discovery-only fallbacks—not acceptance coverage.

Reviewable .mcpr suites name intent clearly, bind assertions to approved MCP contracts, keep secrets out of suite text, and require human approval before execution. Generated packages stay review-only until approved.

Passing MCP Rigor tests do not authorize a release. A successful run is not automatic release certification. Go/no-go can still block on coverage gaps, freshness, approvals, severity, or discovery-only suites.

MCP Rigor discovery smoke confirms the server can list tools. Acceptance coverage requires approved tool, resource, or prompt contracts and human-reviewed assertions. A successful run does not authorize a release.

Green CI is not release evidence by itself. Preserve traceability from requirement through reviewed tests and governed runs into the Release Compliance Dossier—so decisions cite the intended build.

Intent-based testing starts from the requirement—not a blank suite. Generate a review-only package, approve what humans trust, then run governed automation that preserves evidence for release review.