agentleFS
postzen-devpostzen-dev/instagram-agent-skill/skills/ig-repurpose/SKILL.md

ig-repurpose

Skill34 starsChanged 29 days agoMITView on GitHub

Turn one long asset - a YouTube video, podcast, livestream, newsletter, blog post or client call - into a week of reels and carousels. Use when the user says "repurpose this", "turn this into reels", "I have a video/podcast/ transcript", "cut this up", or pastes something long and wants it on Instagram.

8 other projects use a version of this, 6 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 5 are common, 1 is distinctive, and 1 is very specific to this project.

  • 5

    Common

    Most of the 8 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Common:

    If the user gives a URL and this session has a transcript tool, use it.

    line 19
  2. Common:

    If the source is a video the user owns, ask for the file too.

    line 22
  3. Common:

    Not everything is a Reel.

    line 45
  4. Common:

    If the source is the user's own video, use the actual footage.

    line 61
  5. Common:

    Do not dump four finished scripts at once.

    line 87
  6. Very specific:

    When the posts are shot and cut and the PostZen MCP tools are in this session, offer to schedule the week through /ig-publish.

    line 92
  7. Distinctive:

    If anything in the list changes after the yes, the yes is spent.

    line 113

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

ig-repurpose

One good long asset contains four to six posts. Most people extract one and throw the rest away.

Input

A transcript, an article, a newsletter, a script, a call summary, a livestream. If the user gives a URL and this session has a transcript tool, use it; otherwise ask them to paste. Read the whole thing before extracting anything.

If the source is a video the user owns, ask for the file too. A Reel built on their own footage beats one built on their own words read again.

Extract, do not summarise

A summary of a video is not a Reel. Nobody wants the summary. Go through the asset and pull out the things that stand alone:

pullwhat it is
Claimsevery sentence that would start an argument
Numbersevery figure, cost, duration, percentage
Storiesevery moment with a person, a scene and a cost
Mechanismsevery "the way this actually works is..."
Mistakesevery admission of something that went wrong
Linesevery sentence already quotable as-is

List what you found, with counts, before writing anything. If the asset yields fewer than four items, it is thin, and four posts squeezed out of it will be thin too. Say that.

Then pick the format per extract

Not everything is a Reel.

  • Claim, mistake, story to a Reel. They need a voice and a face.
  • Mechanism, numbered list to a carousel. They need to be re-read.
  • A quotable line to a story frame, not a post.

Then build the week

Each extract becomes one post and each post stands completely on its own. The viewer has not seen the source and never will. Never write "as I said in my latest video". The post is the thing.

Assign a hook formula from ig-reel/hooks.json to each and vary them. Five posts from one source with the same hook shape reads as a content mill, because it is one.

If the source is the user's own video, use the actual footage. The clip where they said the thing, with the real reaction in it, beats a re-record every time. Cut on the sentence, not on the breath.

Order across the week so the strongest claim goes first, the story lands midweek, and the mechanism goes last, when the people who liked the earlier ones are watching for it.

Output

SOURCE: "Why we killed discovery calls" (42 min podcast, 8,900 words)

FOUND  5 claims, 9 numbers, 3 stories, 4 mechanisms, 2 mistakes, 7 quotable lines

WEEK
TUE  REEL      #2  Negative Command  Stop running discovery calls
                                     use the 14:20 clip, he laughs at the end
WED  CAROUSEL  Job B caption         The 4-question form that replaced the call
FRI  REEL      #21 Mid-Sentence      "...and he asked for a refund nine days later"
SUN  REEL      #5  Time Collapse     Six hours a week back, one deleted link

Say "write Tuesday" and I will draft it.

Then draft on request, one at a time, each through /ig-reel and /ig-human. Do not dump four finished scripts at once. They will all sound the same and the user will shoot none of them.

Scheduling the week through PostZen

When the posts are shot and cut and the PostZen MCP tools are in this session, offer to schedule the week through /ig-publish. Two ways to place them:

  • Explicit times. Each post gets its own scheduledFor with the user's timezone offset, straight off the plan above.
  • The queue. Each post goes in with queuedFromProfile, and PostZen gives it the profile's next free slot. This needs a queue with slots, which /ig-plan sets up. Never fetch a slot with getNextQueueSlot and pass it back as scheduledFor.

One confirmation covers the whole list, and it lists everything: each post's day and time with timezone, format, account handle, first line of the caption and media file. In queue mode the time is not fixed until the call returns, so list the slots previewQueue({ profileId }) shows, say that each post takes the next free one at send time, and report the scheduledFor that comes back. Show it, get the yes, then /ig-publish sends them one call at a time and checks each platforms[].status before moving on. Reels can sit in publishing for a while on their scheduled day; that is Instagram processing, not a failure, and it is why the log line carries the PostZen post id.

If anything in the list changes after the yes, the yes is spent. Show the list again.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.