# RP004 — Preparation review and next execution stage

2026-10-11. Explicit internal self-review of operational protocol v0.1. This is preparation readiness review, not package disposition, an external review or a recorded research audit.

## Readiness assessment

The supplement preserves the frozen G0–G4/R0–R3 baseline and PC01/PC02. It defines construction, strict ancestry, root sets, consumption and the limited two-input merge policy. G1 uses two copies of a surviving original; G2 has two symmetric fork outputs. The auxiliary C4 control distinguishes a merge from a one-parent continuation after the R1 projection drops an edge.

Truth is derived from executable construction logs. Six inquiry families have exact arities, answer types and a separate representation-only decoder. UNKNOWN is explicit. The R1 decoder deliberately abstains on some recoverable answers, so its answer rate is not described as representation capacity. R2 retains full topology but cannot always reconstruct operation types or input availability. R3 is sufficient for the declared queries; event grouping, chronology, ownership and content-contribution weights remain outside its contract.

Preparation validation checks six disclosed constructions, 54 truth controls, 216 decoder cells and three same-projection/different-truth controls. Adversarial unit checks cover second-parent loss, symmetric outputs, equal independent states, consumed/future inputs, duplicate outputs/parents, copy-state corruption, state-changing continuation, opaque renaming, projection leakage and corrupted truth/decoder expectations. These engineered expected cases are calibration, not independent evidence or newly discovered outcomes.

The targeted collision review inspects W3C provenance specifications and official Git documentation. It acknowledges familiar DAG/provenance ideas and disclaims novelty and metaphysical interpretation. The controlled entity universe, label-sensitive R1 parent selection and conservative abstention policy are explicit limitations, not hidden extra information.

## Remaining execution work

Implement the bounded 864-cell audit and independent replay, then freeze the reviewed source commit and exact input hashes before running. Independent replay must rebuild log-derived ancestry, roots and consumption without importing the producer's truth/ancestry code. Save complete coverage, denominators and collision certificates. Stop and preserve failures according to the protocol. No output from preparation may be presented as a completed audit.

The preparation is sufficient to build that finite reproducibility audit. Outcomes of these controls are disclosed in advance, and the next audit is a deterministic verification, not statistical confirmation. Do not infer broad representation necessity from a conservative decoder's UNKNOWN values.

## Status and verification

Required checks: `python3 scripts/check-rp004-preparation.py`, adversarial unit checks, frozen baseline integrity, typed-record schema, generated English readers, source download fidelity and internal links. CI repeats those checks on the reviewed commit.

RP-004 remains PLANNED. No RP004 Test, Result or broad Claim has been created. Q-0 and Q-4 remain ACTIVE. RP002A and RP003 remain CLOSED under D-0004 and D-0005. Cycle 01 remains ACTIVE with exactly 2/13 resolved obligations (15.4%). Protocol preparation does not resolve the RP004 package-disposition obligation.

Read the [operational protocol](https://autorite.net/sources/rp004-operational-protocol/) and [prior-art review](https://autorite.net/sources/rp004-prior-art/).