Introduction: Infants with asymmetric brain injury (asymBI) are at high risk of Unilateral Cerebral Palsy (UCP). The Grasp and Reach Assessment of Brisbane (GRAB) was developed to detect asymmetries in unimanual/bimanual upper limb (UL) reach and grasp behaviours in infants with asymBI. This study reports the development of the GRAB and evaluates its construct validity and internal consistency. Material and methods: Prospective study of twenty four infants with asymBI and twenty typically developing (TD) infants at 18 weeks corrected age (C.A.) in a structured play session. Three different coloured toys were presented at the midline in a block design of six 30-s trials of toy presentation, separated by five 30-s trials of no toy presentation. The number and duration of: (i) unimanual contacts; (ii) unimanual grasps; (iii) bimanual midline grasps; and (iv) duration of other unimanual behaviours (e.g. prehensile movements and transport phase) were measured. An Asymmetry Index (AI) was calculated to determine asymmetries between ULs. Possible AI values ranged from 0 to 100%, indicating proportion of toy presentation time that unimanual behaviours were asymmetric between ULs. Internal consistency of both the Time Phase (TP) and Toy Colour Phase (TCP) test items were determined by calculating Cronbach’s alpha coefficients. Each assessment occasion was split into six TPs and two TCPs; whereby one TP comprised one 30-s trial of one toy presentation and one TCP comprised two 30-s trials of the same toy presentation. For TP, seven out of nine unimanual behaviours and two out of three bimanual behaviours demonstrated strong internal consistency (Cronbach’s alpha coefficients 0.72–0.89). No unimanual activity demonstrated the strongest IC (0.89). For TCP, six out of nine unimanual behaviours demonstrated strong IC (0.73–0.82). Number of unimanual contacts and duration of unimanual prehensile movements demonstrated the strongest IC (0.82). Duration of unimanual contribution to hands at midline and duration of bimanual midline behaviour demonstrated the weakest IC for both TP and TCP (0.46–0.50). For unimanual contacts, the asymBI group were more asymmetric between ULs (mean AI = 50%) compared to the TD group (mean AI = 30%). For unimanual grasps, both groups were similarly asymmetric (both mean AI = 40%). The TD group were almost twice as likely to demonstrate bimanual grasps as the asymBI group (incidence rate ratio IRR 1.9, 95% CI 1.4 to 2.5, p < 0.001). Infants with asymBI were less likely to use the impaired UL compared to the unimpaired UL for grasping (IRR 0.6, 95% CI 0.5 to 0.8, p < 0.001); and used the impaired UL for a shorter proportion of time compared to the unimpaired UL for grasping (mean difference −9.1%, 95% CI −16.6 to −1.7, p = 0.02). Conclusions: The GRAB is a criterion-referenced research measure that detects and quantifies the presence or absence of unimanual and bimanual reach and grasp behaviours at 18 weeks C.A. in infants at risk of UCP. The GRAB demonstrated moderate to strong construct validity and strong IC within an assessment occasion. There was no toy preference or warm-up effect for TP or TCP for either group; confirming that the GRAB is a consistent measure across toy presentations within an assessment occasion. In this study, the GRAB identified thatinfants with asymBI demonstrated a paucity of bimanual grasping compared to TD infants; and demonstrated asymmetric unimanual grasping between ULs at 18 weeks C.A.