Galaxy clusters and groups are thought to accrete material along the preferred direction of cosmic filaments. Yet these structures have proven difficult to detect due to their low contrast with few studies focusing on cluster infall regions. In this work, we detected cosmic filaments around galaxy clusters using photometric redshifts in the range 0.15<z<0.7. We characterised galaxy populations in these structures to study the influence of "pre-processing" by cosmic filaments and galaxy groups on star-formation quenching. The cosmic filament detection was performed using the AMASCFI Canada-France-Hawaii Telescope Legacy Survey (CFHTLS) T0007 cluster sample (Sarron et al. 2018). The filament reconstruction was done with the DISPERSE algorithm in photometric redshift slices. We showed that this reconstruction is reliable for a CFHTLS-like survey at 0.15<z<0.7 using a mock galaxy catalogue. We split our galaxy catalogue in two populations (passive and star-forming) using the LePhare SED fitting algorithm and worked with two redshift bins (0.15<z<=0.4 and 0.4<z<0.7). We showed that the AMASCFI cluster connectivity (i.e. the number of filaments connecting to a cluster) increases with cluster mass M200. Filament galaxies outside R200 are found to be closer to clusters at low redshift, whatever the galaxy type. Passive galaxies in filaments are closer to clusters than star-forming galaxies in the low redshift bin only. The passive fraction of galaxies decreases with increasing clustercentric distance up to d~5 cMpc. Galaxy groups/clusters that are not located at nodes of our reconstruction are mainly found inside cosmic filaments. These results give clues for "pre-processing" in cosmic filaments, that could be due to smaller galaxy groups. This trend could be further explored by applying this method to larger photometric surveys such as HSC-SPP or Euclid.