Back to catalog / Agent workflows

Read small print from a product photo and retain evidence

Crop and zoom a small label, extract a lot number or warning, and retain the crop.

Goal

Crop and zoom a small label, extract a lot number or warning, and retain the crop.

Submit the image, locate the text, and call the crop tool. [source09]
Read the crop again and archive the original, coordinates, crop, and text. [source09]
Check ambiguous characters manually; do not use low-confidence text for recalls or compliance. [source09]

Verify crop coordinates point to the small-print region and the second read matches original/crop evidence.

If print remains unclear, enlarge the crop without distortion; do not use low-confidence text for recall or compliance.

Anthropic vision API, Python, authorized product photo; no consumer personal data.

Uploads approved product images to Anthropic vision and uses a crop tool; API use may be billed. Retain original, coordinates, and crop under policy.

Crop and text-read results need human confirmation; do not base recall or safety conclusions on low-confidence text.

These values describe the upstream repository, not popularity or likes for an individual skill or playbook.

No metrics snapshot is available yet.

License: MIT