A screenshots folder quickly turns into a digital junk drawer: a receipt, a console error, a recipe, an important screen from an app. Build a script that sorts copies of the files into folders and gives them clear names.
1. Create the folders inbox, sorted and review. Work with copies: the first version of an automation shouldn’t delete originals.
2. The script takes an image and sends it to the model with an instruction to determine the type: receipt, error, note, other.
3. Ask it to answer in one line in the format category|short-name|confidence. For example: error|python-import-error|high.
4. With high confidence, copy the file to the right folder and rename it. Send everything doubtful to review.
The key detail here isn’t image recognition but the right to be wrong. The script doesn’t “tidy up”; it suggests order where it’s confident. That makes the automation useful rather than dangerous.