A Survey-Informed Vision of Open-Source Agents: From Rights to Handoffable Maintenance

Open source gives recipients rights to use, study, modify, and redistribute software, but language-model agents distribute behavior across code, learned parameters, prompts, tools, retrieval, configuration, and runtime state. These mechanisms complicate maintenance in open and proprietary systems alike. Public delivery adds a distinct question: whether recipients obtain enough rights, materials, checks, and modification entry points to exercise maintenance capability and pass it onward. We call this objective handoffable maintenance: a verified maintenance result produced by one recipient becomes a practical starting point from which a subsequent recipient can complete a predefined bounded change. This survey-informed vision separates release-side readiness from demonstrated handoff and organizes prior work through eight maintenance-operation families, artifact roles, preservation relations, evidence, and recipient conditions. It identifies four research pathways: independently usable modification sources; public repair resources derived from restricted failures; propagation of public repairs through customized forks; and recipient-side replacement of propagated unreliable contributions. Each pathway specifies its deliverable, preservation requirements, strong comparisons, and falsifiable prediction. A shared evaluation design distinguishes a full-support reference, independent reception, and onward handoff, with matched information and task-bounded success criteria. Two public industrial change records illustrate how documentary evidence can motivate checkable requirements; this paper reports no agent, CAD, CFD, participant, or cross-recipient execution. Its contribution is a technical synthesis and research agenda for testing when open rights become maintenance outcomes that another recipient can edit, verify, and continue within declared bounds.