Investigate-Consolidate-Exploit (ICE) is a novel strategy for enhancing the adaptability and flexibility of AI agents through inter-task self-evolution. Unlike existing methods focused on intra-task learning, ICE promotes the transfer of knowledge between tasks, similar to human experience learning. The strategy dynamically investigates planning and execution trajectories, consolidates them into simplified workflows and pipelines, and exploits them for improved task execution. Experiments on the XAgent framework show that ICE reduces API calls by up to 80% and significantly decreases the demand for the model's capability. When combined with GPT-3.5, ICE's performance matches that of raw GPT-4 across various agent tasks. ICE enables agents to autonomously adapt and improve performance over time by generalizing past knowledge to tackle new challenges. The strategy disentangles task planning records and execution trajectories as past experiences for inter-task knowledge transfer, promoting efficiency and effectiveness in handling new tasks. The ICE strategy consists of three stages: Investigate, Consolidate, and Exploit. During Investigate, the agent tracks plan and execution trajectories, identifies valuable experiences, and records them. During Consolidate, the agent standardizes the format of experiences for future re-utilization. During Exploit, the agent retrieves and utilizes previously consolidated experiences to improve task execution. The experiments demonstrate that ICE significantly improves task execution efficiency and effectiveness, reduces model API calls, and lowers the barrier for agent deployment. The strategy contributes to a more robust AI community and ecosystem, moving closer to full autonomy.Investigate-Consolidate-Exploit (ICE) is a novel strategy for enhancing the adaptability and flexibility of AI agents through inter-task self-evolution. Unlike existing methods focused on intra-task learning, ICE promotes the transfer of knowledge between tasks, similar to human experience learning. The strategy dynamically investigates planning and execution trajectories, consolidates them into simplified workflows and pipelines, and exploits them for improved task execution. Experiments on the XAgent framework show that ICE reduces API calls by up to 80% and significantly decreases the demand for the model's capability. When combined with GPT-3.5, ICE's performance matches that of raw GPT-4 across various agent tasks. ICE enables agents to autonomously adapt and improve performance over time by generalizing past knowledge to tackle new challenges. The strategy disentangles task planning records and execution trajectories as past experiences for inter-task knowledge transfer, promoting efficiency and effectiveness in handling new tasks. The ICE strategy consists of three stages: Investigate, Consolidate, and Exploit. During Investigate, the agent tracks plan and execution trajectories, identifies valuable experiences, and records them. During Consolidate, the agent standardizes the format of experiences for future re-utilization. During Exploit, the agent retrieves and utilizes previously consolidated experiences to improve task execution. The experiments demonstrate that ICE significantly improves task execution efficiency and effectiveness, reduces model API calls, and lowers the barrier for agent deployment. The strategy contributes to a more robust AI community and ecosystem, moving closer to full autonomy.