Inspiration
I edit and watch a lot of footage, and I kept losing time scrubbing through unsorted clips to find one moment. I wanted a tool that does the sorting and timing for me, so I could spend that time editing instead.
What it does
It takes a folder of raw video clips on a UGREEN NAS, sorts the .mp4 and .mov files by their recorded time, shows me a rename plan, and waits for my approval. After I say yes, it renames the clips in order (for example, BaliVlog-video-0001.mp4), records each rename in a log, and writes a timeline file with each clip’s start time, end time, and new name.
How I built it
I built it as a Skill made of one instruction file, SKILL.md. It delegates the file work to the NAS file agent, using the ug-filesystem and ug-filemgr capabilities. I checked the file with the UGREEN static checker before submitting.
Challenges I ran into
The hardest part was making the Skill safe. A rename can break links in an editing project, and a file can be overwritten by accident, so every step had to wait for approval and stop on any name conflict. I also ran into the checker, which rejected extra files in the Skill folder and wanted every resource linked from the instructions.
Accomplishments that I am proud of
The Skill passes the static checker with no errors and no warnings. More than that, I made it stop instead of guessing: if a file is unreadable, a name already exists, or I haven’t said yes, it does nothing to my footage.
What I learned
Writing instructions for an AI agent is different from writing code. A small wording gap, like not saying what each step returns, can make the whole workflow unclear, and a checker can catch it.
What’s next for Timeline skill
The big open question is whether the NAS reads the recorded dates and durations the way the catalog says. I want to test it on the device, then add a way to describe what happens in each clip if the platform supports it.
Built With
- file-agent
- markdown
- ugos
- uliyaai
