Research questionHow can agent harnesses adapt across tasks and models without manual redesign?Agent harnesses coordinate memory, planning, actions, and tools, but they are often manually designed for individual tasks. This makes it difficult to reuse them across models or adapt them reliably as task requirements change. Latest papersRecent research connected to this question, newest first.JIT-Agent: Scaling Harness Intelligence via Just-in-Time Harness EvolutionThe source concerns machine-generatable harnesses for off-the-shelf agentic LLMs under a fixed four-module protocol covering memory management, planning, action protocols, and tool or skill orchestration. It reports task-time customization, harness repair, and self-evolution, with controlled evaluations on DeepSearchQA and OdysseyBench and comparisons across DeepSeek V4, Mimo-V2.5, Qwen3.6, OpenCode, and Claude Code.research paper · Sep 3, 2026