Brief-to-Footage

Match a client brief or shot list against your NAS photo and video library: see which shots you already own, which are missing, and the best files for each.

Comment

Inspiration

Photographers and editors keep years of work on a NAS, but they can only search it one subject at a time. So they reshoot material they already own, or find out during the edit that a required shot is missing, when a reshoot is expensive. A brief is a list of shots; the archive should be checked against that list.

What it does

Brief-to-Footage is a Skill for the UGREEN AI NAS assistant. You give it a client brief, script or shot list. It:

  1. extracts the shot list and asks you to confirm it
  2. reads the labels your photo library really has (objects, people, places)
  3. searches candidates for each shot by label, folder, date, media type and lens
  4. checks the image candidates visually and gives each a verdict with a one-sentence reason
  5. shows a coverage report with one status per shot: Covered, Partial, Unverified or Missing

If you agree, it also creates a pull-list album and saves the report.

It works at two moments: before a shoot (search the whole archive, shoot only what is missing) and after a shoot (search the shoot folder, catch missed shots while a reshoot is still possible).

How I built it

The Skill is Markdown only: one SKILL.md and four reference files. It delegates each step to a NAS expert agent (file, document, photo, knowledge) and passes only real results from one step to the next. I built it against the official uliya-skill-kit 0.7.3 and its capability catalog, and it passes the kit's static checker with zero errors and zero warnings.

Challenges

  • The official template does not pass the official checker, so I read the checker's rules directly.
  • Object and place search only works with labels the library already has, so the Skill reads the real labels first.
  • The visual check is the least proven step. The Skill separates search from visual checking, so a failure there still leaves useful results, clearly marked Unverified.

What I learned

An honest "Missing" is more useful than a look-alike result. The Skill never widens the search or substitutes similar files on its own.

What's next

Testing on a real UGREEN AI NAS: label listing, multi-image visual checks, speed, and album and file name collisions. The Skill has not been run on a NAS yet. Video candidates are listed but not checked visually.

Built With

  • claude
  • markdown
  • ugos
  • ugreen-ai-nas
  • uliya-skill-kit
Share this project:

Updates

Submission history