# Research execution readiness

These are execution gates for living protocol supplements. The baseline remains unchanged. Bootstrap may create skeletons with explicit pending fields; it must not present them as completed preregistrations or fabricate results. Research execution still requires bootstrap approval.

## Common gate
Before a package's confirmatory runs, commit its prior-art collision record, protocol version and hash, exact question/scope, operational definitions, generators or case fixtures, measures, comparison procedures, resource budget, reproducibility environment, acceptance checks, stopping rules, failure handling and interpretation boundaries. Declare exploratory work separately. Any scientific definition change requires an explicit baseline amendment. Freeze the supplement before observing confirmatory outcomes; later deviations preserve the old version and state whether results were already observed.

## RP002A
Specify the predictive query and horizon; E0/E1/E2 generator parameters and sanity checks; episode counts/lengths; independent train/validation/test splits and leakage checks; complete seed schedule; B0–B3 architectures, capacity/resource controls and optimization budgets; raw-history length and recurrent-state size; the selected primary metric (log loss is currently only a candidate); aggregation unit; effect-size threshold; uncertainty method; calibration diagnostics; treatment of failed runs; multiple-comparison rules; and the operational meaning of “not systematically” and B1≈B2. Choose these before confirmatory outcomes, with a rationale or a separately declared pilot. No numerical defaults are adopted by this handoff.

## RP003
Define the preservation predicate, tolerance and composition/path semantics; deterministic P0–P3 fixtures; expected positive-control behavior; counterexample verification; and the boundary between enumerated examples and general proofs. Retain the non-metaphysical interpretation lock.

## RP004
Specify typed G0–G4 fixtures, ground-truth lineage relations derived from construction, R0–R3 encodings, the exact queries each encoding must answer, scoring/verification and merge semantics. Human intuition remains outside the ground-truth contract.

## RP005A
Specify graph/target construction, restriction rules, planner tie-breaking and horizons, viability criterion, V0–V2 fixtures, seeds, coverage denominators, primary/secondary measure calculations and uncertainty/decision procedures. Verify the structural reachability inclusion independently of the simulation benefit question.

## DAM X
Require supplied source documents with provenance and a bounded pilot question. Until then, preserve only the template and a documented source dependency; do not reconstruct content from assumptions.

## Current readiness
All five workstreams remain pending execution readiness. Missing experimental parameters are deliberate open research choices, not validator defaults. Bootstrap completion, protocol readiness and authorization to execute are separate checks.