tagging-skill scans images, stores categorized tags and descriptions, and opens a local page where you can search, filter, remove, restore and add tags.
Your original images stay where they are. The Skill drafts the metadata; you decide which tags to keep.
The run included photographs, an illustration and composite imagery. Completion counts do not measure tag accuracy.
This run checked the result grid, individual details and keyword search after tagging.
BeforeA cherry photo sits somewhere in a project or date folder. Finding it requires remembering its location or opening files one by one.
AfterThe Skill stored tags including cherry and water droplet. Searching cherry in the review page returned the image.
Visible content becomes searchable text.
BeforeAn automatic tag may not fit your judgment, and a one-off deletion may be lost during another analysis.
AfterYou can remove, restore and add tags. A removed tag leaves a rejection record that subsequent reanalysis respects.
Human decisions stay with the library.
The interaction screenshot uses separately prepared demonstration data.
BeforeAs new photos arrive, you have to work out which files were already processed.
AfterA normal rerun skips unchanged completed images. Rescanning this run found zero added, changed or removed images and no pending work.
Existing results carry forward.
Incremental handling of newly added or changed files was not separately demonstrated here.
Finding and reusing assets falls back to memory and manual browsing.
Photographers and content teams often organize images by project or date. When they need a subject, scene or visual detail, filenames rarely answer the question. Writing keywords for every image by hand is difficult to maintain.
A full asset manager can add another import and maintenance workflow. A one-time tagging pass still leaves incorrect labels and future additions to manage. For a folder-based library, a searchable, editable metadata layer is a direct way forward.
Keep the photos in place. Make the tags open to correction.
tagging-skill builds metadata beside a chosen folder and presents each result for human review. The Agent produces a first pass; the user makes the final call.
Provide a folder containing JPG, JPEG, PNG or WebP images, including any subfolders. Original files remain in place.
A vision-capable Agent writes a description and tags under subject, scene, activity, visual, condition and mood.
Filter by status, search stored text, inspect descriptions and edit tags. Run the Skill again as the folder changes.
The tag database lives inside the photo folder. The local review page does not run AI tagging. Image analysis is performed by the Agent running the Skill, and whether that involves a remote model depends on the host environment.
Initial tags in these cases came from the actual vision pass and were not manually seeded.
The image shows a red cherry with droplets, with another cherry blurred behind it. The Skill produced four tags:
“Water droplets hang from a red cherry with another cherry blurred behind it.”
Searching cherry in the review page returned this photo.
Subject, action and visual overlay entered separate tag categories, giving multiple ways to locate the same image.
The Skill also produced reviewable tags for an illustration in a supported format. This does not demonstrate performance on real cat photos.
All nine images completed with zero failed items. Displayed confidence values are model-reported evidence strength, not measured accuracy. This run did not measure time saved.
Original images stay in place, while the database is stored inside the chosen folder. Start with the assets you already have.
Inspect descriptions and categorized tags for each image, then remove, restore or add tags as needed.
Removing a tag creates a rejection record. Later analysis of that image respects the recorded decision.
Normal reruns prioritize new, changed or previously failed images. The unchanged images in this run produced no pending work on rescan.
As a collection grows, it is easy to lose track of what each image contains and why a label was kept or rejected. tagging-skill lets new images join the same metadata library and lets each review decision inform later work.
Finding an image need not start with opening every file again.
Provide the absolute path to your photo folder. The Skill scans supported images, writes tags and descriptions, and opens a local page where you can inspect, remove, restore and add tags.