Remix Kit: New Content From Your Archive

Finds your notes, photos and recordings on a topic across your NAS and turns them into a blog post, newsletter, script or carousel. Every claim cited, nothing invented.

Comment

Inspiration

Years of notes, screenshots, photos and recordings pile up on a NAS, yet every new post still starts from a blank page. The material for a great newsletter or video often already exists. It's just buried, and when I do dig something up, I keep reusing the same few assets. I wanted a Skill that treats the archive as raw material.

What it does

Remix Kit takes a topic and a format and builds new content from what you already have:

  • Searches the archive by topic across documents and notes, photos and screenshots, and recordings.
  • Shows a shortlist so you choose what goes in.
  • Transcribes chosen recordings with timestamps.
  • Writes the piece in one of eight formats: blog post, newsletter, short video script, video outline, carousel, thread, case study or FAQ.
  • Traces every claim and visual to a source, and adds a "Gaps" note where the archive has nothing, instead of inventing.
  • Saves the piece after approval, optionally builds an album, and tags used sources remix-used so the next run prefers fresh material.

How we built it

It's a Markdown-only orchestration Skill written to the Uliya Skill Development Standard, plus a formats reference:

  1. knowledge-agent: evidence search across documents.
  2. photo-agent and file-agent: visual search with captions and relevance scores, recordings, and the reuse ledger.
  3. User shortlist.
  4. knowledge-agent: transcription.
  5. document-agent: writing the piece with sources.
  6. file-agent and knowledge-agent: saving it.
  7. photo-agent and file-agent: album and tags.

I used Claude Code to read the Skill Developer Package and capability catalog, draft the Skill and format templates, and run the kit's static checker.

Challenges we ran into

  • Searching three very different source types and merging them into one shortlist without flooding the user.
  • Recordings have no content index before transcription, so they're found by file and folder name and only transcribed once chosen, to keep runs fast.
  • Making "nothing invented" enforceable: every format template ends with Sources and Gaps.
  • The static checker expected different headings and evidence formatting than the kit's template; I read its source to pass cleanly.

Accomplishments that we're proud of

  • Passes the official static checker with zero errors and zero warnings.
  • The remix-used reuse ledger: a simple NAS tag that changes which material later runs pick.
  • One Skill that covers eight content formats through a single editable reference file.

What we learned

  • A human shortlist step between search and generation makes results much better and keeps the creator in control.
  • NAS tags work well as lightweight memory between Skill runs.

What's next for Remix Kit: New Content From Your Archive

  • Validate search recall and citation accuracy on a real UGREEN AI NAS.
  • Series mode: plan several pieces from one topic without repeating sources.
  • Chain with Crosspost Kit to turn each piece into a ready-to-post pack.

Built With

  • ai-agents
  • claude-code
  • markdown
  • semantic-search
  • skills-hub
  • speech-to-text
  • ugos
  • ugreen-ai-nas
  • uliya
  • vision-language-model
Share this project:

Updates

Submission history