Someone should have to prove it.
Models are asked to write the same code over and over, and they get the same things wrong over and over. Not the hard things: the exact ones. A page that has to be landscape. A form field named by a tax agency rather than by its label. A chart the spreadsheet quietly discards because an axis was not declared.
A model can be told about those a thousand times and still miss them, because it is writing the code fresh each time. So write it once, have a program check it against cases it has never seen, and let every model call that instead. That is the whole idea, and it is not complicated. What is unusual is insisting on the check.
The check is why a skill here is worth something. Anyone can publish code. Almost nobody will tell you what it scored, on what, measured how, and what it does not do. Every skill on this shelf carries that, and it is re-run on fresh cases rather than signed once and forgotten.
What this refuses to do
It will not list a skill that cannot be checked. If no program can say whether the output is right, it does not go on the shelf, however useful it sounds.
It will not build one that is not needed. Of thirteen kinds of job we measured recently, a cheap model already handled nine, so nothing was built for them. One skill was built, measured and thrown away because its gain sat inside the noise.
It will not charge for access. Reading, installing and publishing are free at any volume. Money comes from work done for you: building a skill on your own failing task, private shelves, higher limits, and a fee on paid bounties.
It will not hold your money. A bounty is paid by the person who wanted the work to the person who did it. We are not an escrow agent and will not pretend to be one.
Who
Clouthier Simulation Labs, in Ontario, Canada. Small enough that when you write, you get a reply from the person who wrote the code.
The loop that builds these is open source under Apache 2.0. You can run the whole thing on your own machine with your own keys, and keep what it makes.