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 and no interface file or helper library beside it: a small state machine with a deliberately tiny state space — a handful of external functions, a handful of distinct states, no owner, no privileged caller, no upgrade path, no handling of funds — and an explicit custom error for every input it refuses. The advance is in what the checking can claim rather than in how hard it pushes.

Everything this tree has accepted was checked by sampling: randomised sequences against a naive model, invariants asserted along the way. That says the suite looked hard, not that it looked everywhere.

Here the test tree must instead enumerate exhaustively: starting from the initial state, apply every available call with every representative argument to every reachable state, out to a stated depth, deduplicating states you have already visited, and assert the full invariant set at each one — including that the contract's refusals are exactly the transitions the specification forbids, no more and no fewer.

Choose the state space small enough that the enumeration completes in the suite, and say in the project's documentation what the depth bound is, how many distinct states were visited, and which behaviours lie outside the bound and are therefore not covered by the complete claim. A test that quietly explores less than it says it does is worse than one that samples honestly.

Finally, keep the mutation discipline that has proved its worth: inject faults into the contract one at a time, record which the enumeration catches, and include at least one fault your first suite genuinely missed — written down as a miss, before the strengthening that caught it. Everything must build and run offline from the committed files alone.