Drawing on the Sutra of Perfect Enlightenment (Yuanjuejing) and repentance liturgies such as Zongmi's Manual for Cultivating Realization in the Place of Practice of the Sutra of Perfect Enlightenment (Yuanjuejing daochang xiuzheng yi), this essay examines the visual program of the Southern Song–dynasty Cave of Perfect Enlightenment (Yuanjuedong) at Baodingshan, Dazu. It argues that the cave serves not as an arena for ritual practice per se but instead presents an idealized representation of such a site. In particular, this essay focuses on the conceptual function of the sculpture of the kneeling bodhisattva in the middle of the cave, a figure that finds no exact counterpart in scriptural sources. Investigating the significance of such figures that mediate between worshippers and their deities in various media in the Song, this essay contends that the kneeling bodhisattva not only enacts perpetual reverence on behalf of absent worshippers but also solicits spectatorial self-identification, enabling viewers to imagine themselves into the representational world of scriptural narrative and liturgical practice constructed in the cave. This essay then interrogates the triangular relationship between illusion, matter, and the mind that is thematized in the cave and in liturgies like that by Zongmi. Engaging with Hans Belting's recent work on image anthropology, this essay concludes by suggesting that the cave, its source texts, and related repentance rituals collectively insist on the fundamental irrelevance of all media save the mind itself. Such a notion points to the need to develop a specifically Buddhist theory of images, media, and minds.