format in medical domains for prescribing a set of rules and policies that an attending physician should follow. In terms of an AI planning task, clinical guidelines can be viewed as a shared library of highly reusable skeletal reactive plans, whose details need to be refined by the executing planner over significant periods of time. The application of clinical guidelines involves the collection and interpretation of patient-related data, the application of prespecified plans, and a revision of the plans when necessary. Over the past decade, several research groups have implemented reactive-planning architectures specific for the task of refining skeletal plans over time. The importance of such systems is increasing as more clinical data are being captured and represented in an electronic format, and as quality control of medical care grows in importance. Conducting a flexible, intelligent dialogue with the physician user of these systems requires an ability to reason about the user’s goals and plans and about possible modifications to these plans. We point out that automated support for guideline-based care can be viewed as a collaborative effort of two planning agents: the physician and an automated ("assistant") planner, whose knowledge might limited, or who might not have access to all the data. We demonstrate that achieving even a modicum of collaboration between the two planners and of highly desirable flexibility in the automated support to the physician involves a form of reasoning about mental states: a recognition of each planner’s (in particular, the physician’s) intentions and plans to achieve them, and a consideration of the available plan-revision strategies during execution time. In particular, automated support for clinical guidelines could be enhanced considerably by a sharable, explicit, formal representation of (1) therapy-planning-operators’ effects, (2) plan-revision strategies, and (3) the underlying goals and policies of the guideline, in the form of temporalabstraction patterns to be maintained, achieved, or avoided. Finally, relying on our analysis, we can list in a structured manner several possible sources of disagreement between the two planners, thus supporting the maintenance of the automated planner’s knowledge base as well as improving the appropriateness, and thus acceptability (by the physician) of its future recommendations.