Focused ultrasound (FUS) has been shown promise in treating the brain locally and noninvasively. Transcranial passive cavitation detection (PCD) provides methodology for monitoring the treatment in real time, but the skull effects remain a major challenge for its translation to the clinic. In this study, we investigated the sensitivity, reliability, and limitations of PCD through primate (macaque and human) skulls in vitro. The results were further correlated with the in vivo macaque studies including the transcranial PCD calibration and real-time monitoring of blood-brain barrier (BBB) opening, with magnetic resonance imaging assessing the opening and safety. The stable cavitation doses using harmonics (SCDh) and ultraharmonics (SCDu), the inertial cavitation dose (ICD), and the cavitation SNR were quantified based on the PCD signals. Results showed that through the macaque skull, the pressure threshold for detecting the SCDh remained the same as without the skull in place, whereas it increased for the SCDu and ICD; through the human skull, it increased for all cavitation doses. The transcranial PCD was found to be reliable both in vitro and in vivo when the transcranial cavitation SNR exceeded the 1-dB detection limit through the in vitro macaque (attenuation: 4.92 dB/mm) and human (attenuation: 7.33 dB/ mm) skull. In addition, using long pulses enabled reliable PCD monitoring and facilitate BBB opening at low pressures. The in vivo results showed that the SCDh became detectable at pressures as low as 100 kPa; the ICD became detectable at 250 kPa, although it could occur at lower pressures; and the SCDu became detectable at 700 kPa and was less reliable at lower pressures. Real-time monitoring of PCD was further implemented during BBB opening, with successful and safe opening achieved at 250 to 600 kPa in both the thalamus and the putamen. In conclusion, this study shows that transcranial PCD in macaques in vitro and in vivo, and in humans in vitro, is reliable by improving the cavitation SNR beyond the 1-dB detection limit.