Progress on NLP for mental health — indeed, for healthcare in general — is hampered by obstacles to shared, community-level access to relevant data. We report on what is, to our knowledge, the first attempt to address this problem in mental health by conducting a shared task using sensitive data in a secure data enclave. Participating teams received access to Twitter posts donated for research, including data from users with and without suicide attempts, and did all work with the dataset entirely within a secure computational environment. We discuss the task, team results, and lessons learned to set the stage for future tasks on sensitive or confidential data.