PromptCIR is a prompt-learning-based blind compressed image restoration (CIR) method that effectively restores images from various compression levels. Unlike traditional methods that rely on predicted quality factors, PromptCIR uses lightweight prompts to implicitly encode compression information, providing dynamic content-aware and distortion-aware guidance for restoration. The method leverages a transformer-based backbone with a dynamic prompt module to handle blind CIR tasks efficiently. PromptCIR utilizes a U-shape structure and incorporates a hybrid attention block (RHAG) to enhance local and global information extraction capabilities. The dynamic prompts are generated using a dynamic prompt module (DPM) that adaptively encodes spatial information. PromptCIR was tested on the NTIRE 2024 challenge and achieved first place in the blind compressed image enhancement track. Extensive experiments demonstrated the effectiveness of PromptCIR in restoring compressed images with high quality. The method is lightweight, with minimal parameter overhead, and is suitable for deployment on edge devices. PromptCIR outperforms existing methods in both quantitative and qualitative evaluations, achieving superior performance in restoring texture details and reducing compression artifacts. The method is also effective in non-blind CIR tasks, showing improved performance with larger training datasets and enhanced representation capabilities. Overall, PromptCIR provides a powerful solution for blind CIR tasks, demonstrating the potential of prompt learning in image restoration.PromptCIR is a prompt-learning-based blind compressed image restoration (CIR) method that effectively restores images from various compression levels. Unlike traditional methods that rely on predicted quality factors, PromptCIR uses lightweight prompts to implicitly encode compression information, providing dynamic content-aware and distortion-aware guidance for restoration. The method leverages a transformer-based backbone with a dynamic prompt module to handle blind CIR tasks efficiently. PromptCIR utilizes a U-shape structure and incorporates a hybrid attention block (RHAG) to enhance local and global information extraction capabilities. The dynamic prompts are generated using a dynamic prompt module (DPM) that adaptively encodes spatial information. PromptCIR was tested on the NTIRE 2024 challenge and achieved first place in the blind compressed image enhancement track. Extensive experiments demonstrated the effectiveness of PromptCIR in restoring compressed images with high quality. The method is lightweight, with minimal parameter overhead, and is suitable for deployment on edge devices. PromptCIR outperforms existing methods in both quantitative and qualitative evaluations, achieving superior performance in restoring texture details and reducing compression artifacts. The method is also effective in non-blind CIR tasks, showing improved performance with larger training datasets and enhanced representation capabilities. Overall, PromptCIR provides a powerful solution for blind CIR tasks, demonstrating the potential of prompt learning in image restoration.