Ready to Build IV: Closing the Transition Function — Fail-Closed Defaults as Conformance Criteria in the Freshness Portability Agent

This short paper reports what a fourth agent adds to a series on execution-neutral specifications. The first paper (https://doi.org/10.5281/zenodo.23229267) presented execution-neutral buildability; the second (https://doi.org/10.5281/zenodo.23342767) argued that a specification can fix which of its outcomes must stay distinct; the third (https://doi.org/10.5281/zenodo.23386040) reported conformance records that assert the emitted verdict together with the state. Each supplementary projection in (https://doi.org/10.5281/zenodo.23342767) and (https://doi.org/10.5281/zenodo.23386040) left a large share of its space without an outcome — 134 of 271 rows and 170 of 192 inputs — and, for the state and trigger pairs the workflow does not enumerate, the reconstructions read the specification as preserving the current state for non-critical triggers in (https://doi.org/10.5281/zenodo.23342767) and as fixing fail-closed semantics in (https://doi.org/10.5281/zenodo.23386040). The Freshness Portability Agent pack (FPA-FRESHNESS-PORTABILITY-003, v1.0) specifies a freshness-governed control for import acceptance, receipt anchoring, watermark advancement, and restoration eligibility. Its conformance suite closes the transition function at the level of state and trigger pairs: a scenario record grounded in the Specification's closed-state rule states that any pair the workflow does not enumerate fails closed to the rejected state, and runnable records assert this default on four representative pairs, with a fifth on the rejected state itself, where failing closed and preserving the state coincide. In the pack's NNT companion run, five paradigm-isolated reconstructions in Python, Rust, Prolog, C, and Bash ran the 41 conformance records live and each reported 35 PASS, 6 SKIP, 0 FAIL, and 0 ERROR, with no divergent record and no corrections; four classified the question of unenumerated pairs as resolved by the pack, and the fifth as a mechanism under constrained semantics. In the recorded mutation run, the mutants that replaced the default with silent preservation of the state changed the status of the four records that assert it on states other than REJECTED, in each of the three reconstructions where they were built. The pack also keeps the canonical verdict, the operational action token, and the gate class on separate layers, and marks two of its 27 generic scenarios NOT-APPLICABLE, each with a stated reason. A supplementary artifact renders a declared 252-input projection in five further paradigms, with byte-identical text outputs and matching spreadsheet results, and assigns an outcome to all but one input, 224 of them through the fail-closed default for unenumerated pairs. The closure has a bound that we report: on six branches of permitted transitions that no record asserts, the five reconstructions and the projection do not all agree. We read this as evidence that a suite can fix, and check on representative pairs, the outcome of the state and trigger pairs its workflow does not enumerate, for this artifact; it is not a certification or a quality score. Use of AI tools. The author constructed the pack's seven canonical files, including the conformance tests, reviewed them manually, and designed and calibrated the harnesses for the reconstruction campaign and the supplementary derivations. Large language models were used openly, as tools to accelerate research and production, with manual review: they assisted the drafting of the pack's documents and the revision of this manuscript and, by design, built the reconstructions and the supplementary derivations that are the object of the demonstration. No tool is an author; the author is accountable for the content. What matters is not which tools touched the documents but whether the artifact can be verified. The claims in this paper are limited to what a reader can check with the shipped hashes, the runnable conformance suite, and the reproducible outcome vectors.