Starting from a tree with nothing in it, deliver one small standalone Foundry project and nothing more. Write the build configuration and the directory layout by hand, and depend on nothing that must be fetched — no package manager, no vendored test framework, no submodules of any kind — because the verifier re-runs the build and the suite offline with no network and no credentials, and a pointer it cannot fetch reads as a build failure rather than as the cause.

The source tree holds exactly one contract: a modest state machine or registry behind a narrow, documented external interface of a handful of functions, declaring an explicit custom error for every input it refuses, with no owner, no privileged caller, no upgrade path and no handling of funds. Do not add a second contract, an interface file, or a helper library beside it.

In the test tree only, write a second and deliberately naive model of the same machine — the obvious, slow, plainly-correct version — and drive long randomised sequences of calls into both the contract and the model, comparing the full observable state, every counter and the exact reason for every refusal after each individual step rather than only at the end. The advance this time is proving the suite can fail.

Alongside the suite, commit a plain document that names each way you deliberately broke the contract to test the harness — an off-by-one in a bound, a skipped state transition, a swapped error, a comparison in the wrong direction, a stale read — and for each one records which test caught it and how it failed. The delivered tree must be green on the honest contract, and the document must show the suite going red on every injected fault.

Keep the whole thing small enough to build and check comfortably offline; the point is a suite of known strength, not a large project.