The closest example I can point to at the moment is the WTO Government Procurement Agreement, plurilateral, heavy on transparency. The empirical work on it finds lower corruption-risk indicators such as fewer single-bidder contracts, fewer wins piling up with one firm. A caveat is that the GPA bundles transparency with non-discrimination and open tendering, so transparency alone doesn't get full credit.
On confidentiality, the objection every disclosure proposal runs into, the Seoul Frontier AI Safety Commitments offer something borrowable. They cover safety-framework implementation rather than contract terms, but the structure is more important here. Disclose publicly, unless doing so increases risk or divulges commercially sensitive information out of proportion to the societal benefit and whatever can't be public still goes confidentially to trusted actors. That fallback is the useful part, I think. It maps neatly onto disclosing to a coalition secretariat instead of every member.
I think the weak point is that the carve-out is self-assessed, so the firm decides what counts as disproportionate. Designing verification around this kind of gap is one of the aspects my current paper is looking into. @atilgandev