Researchers introduce Doc2Harness, a two-stage benchmark that directly evaluates how well AI models can build the software harness surrounding a coding agent, rather than just measuring downstream task success under a fixed harness. In Stage 1, a build model constructs a complete candidate harness from an interface-only contract and a development document consolidating behavior across twelve frontier harnesses; in Stage 2, the frozen harness is driven end to end by a scripted mock model and scored on four observable channels, with positive controls ruling out vacuous passes. Across nineteen frontier models tested on 328 test cases, GPT-6 Astra achieved the highest pass rate at 53.05%, while the weakest model reached only 27.44% — and even the strongest model failed nearly half of all test cases, with failures clustering around how cross-cutting functionality is split across modules.
