Inspiration
Downloads pile up in an incoming folder, whether that’s sample packs, video clips or other audio and video. Sorting them by sample rate, bit depth, codec, duration and resolution by hand is slow and error-prone. A NAS already has the files and the tools to read them, but it has no easy way to turn that into an organized index. I wanted that to be an AI skill you can ask for in plain language, with you staying in control of what gets changed.
What it does
- Scans an incoming directory you name, including subfolders, for audio and video files.
- Reads each file’s technical metadata and groups the files into spec categories, such as hi-res audio, CD-quality audio, 4K, HD and SD video, and short clips.
- Shows an index report with the specs behind every placement, and a separate Unclassified group for files with missing data.
- Proposes a tagging plan and stops for your approval. Only after you confirm does it create tags and apply them to the approved files.
- It never moves, copies, renames or deletes files.
How we built it
It’s a single orchestration Skill for the UGREEN NAS agent platform. The parent Skill only delegates, and a registered file expert does the real work using the NAS file system and file manager capabilities. The flow has three steps: a read-only scan and metadata extraction, a report and plan shown in the chat, then an approval-gated tagging step. We checked every capability against the public catalog and validated the Skill with the kit’s static checker, which passed with 0 errors and 0 warnings.
Challenges we ran into
Our original idea included mood tagging of samples, clips and photos, and automatic ingest. The platform doesn’t support those, because it has no audio or video content analysis and no folder watcher. We narrowed the scope to what the platform can do reliably. The catalog doesn’t document which fields the metadata tool returns. The Skill therefore reports any missing field as “not reported” and never guesses. Metadata is read one file at a time, so the Skill uses a file limit per run.
Accomplishments that we're proud of
Every write action sits behind explicit user approval. It reports honestly when data is missing or a step fails. It passes the platform’s static check.
What we learned
Designing around approval and partial results matters as much as the happy path. Checking what the platform can do before designing the workflow saved us from building something that couldn’t run.
What's next for Media Spec Indexer
Mood tagging for photos using the platform’s vision capabilities. Creating photo albums from the tagged results. Scheduled or triggered runs once the platform supports them. Optionally saving the index report to the NAS.
Built With
- nas
- the-file-agent-expert
- ug-filemgr-(media-info-and-tags)
- ug-filesystem
- ugreen