# Exact unsent Grok prompt — EM-PRD-04

Independent enterprise metamodel review. Do not browse, invent identifiers or claim standards conformance.

EM-PRD-04 covers ResearchStudy, Hypothesis, ExperimentDesign, ExperimentRun and ResearchFinding using WM-ACT-036 Research Study, WM-KNW-009 Hypothesis and reserved-but-spec-missing WM-ACT-022 Experiment / Trial, with claim, evidence, observation and dataset boundaries.

Assess this proposal: reuse WM-ACT-036 and WM-KNW-009; complete WM-ACT-022 narrowly as Experiment Run; study-scoped Experiment Design is an immutable WM-ACT-036 protocol/design release; only a reusable cross-study design is an identifier-unassigned candidate; Research Finding profiles WM-ACT-036 plus WM-KNW-007/008. Keep study, hypothesis, design, run, attempt, observation, dataset, analysis and finding distinct. Preregistered hypotheses precede result access; post-hoc hypotheses are explicit. Run manifests pin exact design, inputs, code, configuration, environment, randomness and references. Missing data, zero, censoring, no effect, null and inconclusive remain distinct. Findings record uncertainty, applicability and limitations.

Test two runs of one design with different input snapshots and contradictory outcomes while retaining failed attempts and the original hypothesis. Reject deleting a failed experiment or rewriting a hypothesis after results.

Return <=900 words with: Verdict; strongest evidence; strongest counterexample; identity/mastership; study/design/run; hypothesis lifecycle; provenance/reproduction; observations/missingness; analysis/finding; negative/null/inconclusive; scenario; at least 10 invariants; minimum model set; blockers. Explicitly decide which candidates require independent identity.