[LorenzoRinaldi]

The existing study starts from prepared mix workbooks and machine-specific paths, making the GCAM-to-MARIO interpretation difficult to follow. This adds an executable route from physical GCAM CSV outputs through explicit technology/region concordances and share normalisation to native MARIO u, s and Yc changes, with actual coefficients read back alongside the expected formula.

The README and executed notebook follow one published wind-generation value (4.51167191 EJ/yr) into a 0.391284374506 share and a MARIO coefficient. Users can inspect inputs without a database, isolate electricity or steel, run both on a small teaching SUT, or apply the same functions to extended EXIOBASE. The teaching inventories/demand are clearly fictional; GCAM inputs are actual study results.

Included inputs and behaviour:

• Electricity physical outputs extracted from the fixed public supporting-information workbook, retaining values, units, scenario/year labels and source provenance.
• Regional steel physical outputs from the maintainer's SteelProdTech export. Their EU-12 + EU-15 sums match all 504 published EU27 scenario/year/route values (maximum absolute difference 1.42e-14 Mt/yr).
• Strict file contracts, visible many-to-one technology mapping, exact zeros, independent scenarios and preservation of electricity totals per supplying region/consumer and selected steel supply totals. The legacy electrolysis exclusion is explicit and configurable.
• Auditable output files: original input row/value/denominator, target mixes, before/after coefficients, footprints and integrity/provenance manifest.
• Pinned MARIO 1.0.0 runtime, fixed v1.3.0 source download with resume/checksums, reusable prepared bundles with a real parser reload before completion, packaging and CI.

Validation: 25 tests passed, all 11 notebook code cells executed, documentation/input hashes checked, and the figure visually inspected. The full 23-region source was prepared, exported and reloaded. An installed wheel then ran the user CLI outside the checkout and offline; independent checks verified the unchanged baseline and satellite coefficients, conserved electricity/supply totals, electrolysis exclusion, output hashes and actual coefficient values. The live fixed-record download/resume path was checked on a small source file; large matrices already verified from the byte-equivalent v1.3.2 record were reused.

The historical study scripts, inventories and derived workbooks are unchanged. This new route applies EU mix shocks, retaining background demand/trade and other regions' mixes. It does not run GCAM or claim full reproduction of the paper's figures or absolute consumption/import volumes. The old driver's unapplied demand/import logic is documented in docs/research-workflow.md rather than silently treated as a working trade shock.

Evidence: docs/verification.md, verification/full-model.json, source receipts and the executed notebooks/GCAM_to_MARIO.ipynb.