The application and analysis of the Cut-and-Choose technique in protocols secure against quantum adversaries is not a straightforward transposition of the classical case, among other reasons due to the difficulty to use rewinding in the quantum realm. We introduce a Quantum Computation Cut-and-Choose (QC-CC) technique which is a generalisation of the classical Cut-and-Choose in order to build quantum protocols secure against quantum covert adversaries. Such adversaries can deviate arbitrarily provided that their deviation is not detected. As an application of the QC-CC we give a protocol for securely performing two-party quantum computation with classical input/output. As basis we use secure delegated quantum computing (Broadbent et al 2009), and in particular the garbled quantum computation of (Kashefi et al 2016) that is secure against only a weak specious adversaries, defined in (Dupuis et al 2010). A unique property of these protocols is the separation between classical and quantum communications and the asymmetry between client and server, which enables us to sidestep the quantum rewinding issues. This opens the prospect of using the QC-CC to other quantum protocols with this separation. In our proof of security we adapt and use (at different parts) two quantum rewinding techniques, namely Watrous' oblivious q-rewinding (Watrous 2009) and Unruh's special q-rewinding (Unruh 2012). Our protocol achieves the same functionality as in previous works (e.g. Dupuis et al 2012), however using the QC-CC technique on the protocol from (Kashefi et al 2016) leads to the following key improvements: (i) only one-way offline quantum communication is necessary , (ii) only one party (server) needs to have involved quantum technological abilities, (iii) only minimal extra cryptographic primitives are required, namely one oblivious transfer for each input bit and quantum-safe commitments.