This qualitative embedded single-case study sought to describe a dance intervention designed for people in stroke rehabilitation. Recommendations from rehabilitation therapists, and the researcher-dance educator's observations of the therapists' work and of patient participation during the intervention were captured and analyzed, from design to implementation, through various techniques, including a validated descriptive checklist from the health field, namely the Template for Intervention Description and Replication (TIDieR). This methodology enabled us to describe what 'adapted' dance was in this context, in terms of content and pedagogy, and thus to better distinguish such an intervention from the psychotherapeutic models underlying dance therapy. Thirteen rehabilitation themes emerged from the data to inform the dance content (the 'what' of the intervention), while four pedagogical categories subdivided into modalities portrayed the 'how'. These findings suggest a 'more experientially based' intervention distinct from 'more functionally based' therapies. This paper presents the perceived effectiveness/ineffectiveness of the pedagogical modalities from the researcher-dance educator's perspective and highlights the contribution of somatic-sensitive pedagogy. It concludes with some limitations of the TIDieR, such as its weakness in providing detailed and nuanced movement descriptions, leading the authors to recommend the use of complementary audiovisual data that would likely facilitate knowledge transfer.