Skip to content
Automation Club
  • MarkdownHow we plan a month of content in one afternoonExample output

    How we plan a month of content in one afternoon

    Source: YouTube auto-captions copied from Show transcript, about 5 minutes, one speaker. [Music] tags removed. No unclear passages.


    Why we stopped posting every day (00:04)

    Hi, welcome back. So today I want to talk about how we plan a month of content in one afternoon, because for a long time we didn't do it that way at all.

    So for about a year we posted every single day. Every single day. And by the end of it, nobody on the team wanted to open the content calendar. Nobody.

    And the posts got worse, not better, which is the opposite of what you'd expect, right? You'd think practice makes it better. But we were making content to fill a slot, not because we had something to say. And people can tell. They can always tell.

    One video a month (00:48)

    So we changed one thing. Just one thing. Now we record one long video at the start of each month, and everything else comes out of that recording. So the short clips, the carousel, the newsletter, the threads, all of it comes from that one video.

    People ask how long the recording takes. It's about an hour, including the bits we cut. And the planning afterwards is one afternoon. One afternoon a month, that's it. That's the whole planning.

    The three parts (01:24)

    So how do we actually do it? There are three parts.

    The first part is before we record. We write down five questions our audience actually asked us that month, in comments, in emails, in DMs, and we answer those on camera. We don't make up topics. We answer real questions.

    The second part, and this is the big one. Say the point in one sentence before you explain it. So every answer starts with the point in one sentence, and then we explain it. And that sentence becomes a post on its own. You can lift it straight out and it works.

    And if you can't do that, if you can't say it in one sentence, the idea usually isn't ready to record yet. So we park it and come back next month.

    The third part is after we record. We clean the transcript before we write anything. So filler out, headings in. And we pull out the best lines, the lines that make sense on their own, and every post is built on one of those lines.

    So the carousel is for saves, the LinkedIn post is for the story and the thread is for the steps. Different shape for each platform, but it's the same idea underneath.

    And then we plan the week in one sitting. One post a day, and one day off on purpose, because you need a day where you're not posting.

    What changed (03:05)

    So what changed? Well, we post less now and we hear back more. More replies, more people saying "I saved this", more people asking follow-up questions, which, by the way, become next month's five questions. So it kind of feeds itself.

    The other thing that changed is the team actually likes it again. Recording day is fun now. It's not a chore. It's one day where we sit down and talk about things we know about.

    Mistakes we made early on (03:40)

    A couple of mistakes we made early on, so you don't have to. The first one was trying to write the posts from memory after watching the video. Don't do that.

    Use the transcript, use the actual words, because the way you say something on camera is usually better than the way you write it later. It's more natural.

    The second mistake was making every post open with the same line, so people who follow us in two places saw the same thing twice. Now no two posts open with the same line.

    And the third one was making the carousel need the caption. A carousel should make sense with no sound and no caption, just the slides. Three slides that work on their own.

    Where to start (04:33)

    So if you want to try this, here's where I'd start. Next time you record anything, a call, a video, a voice note, check every answer for that one sentence. The point in one sentence.

    If it's there, you've got a post. If it's not, the idea isn't ready yet.

    That's it for today. If you have a question about any of this, leave it in the comments, because it might end up in next month's video. See you next time.

    Quotable lines

    1. "Say the point in one sentence before you explain it."
    2. "Now we record one long video at the start of each month, and everything else comes out of that recording."
    3. "Use the transcript, use the actual words, because the way you say something on camera is usually better than the way you write it later."
    4. "A carousel should make sense with no sound and no caption, just the slides."
    5. "Different shape for each platform, but it's the same idea underneath."
    Made with Claude · Skill

    Markdown · Document

    Cleaned transcript. A five minute video's YouTube auto-captions, cleaned. The output is shown as Claude wrote it.

Works with

  • Claude
  • Claude Code
  • ChatGPT
  • +18 more

Not affiliated with or endorsed by the companies shown.

Content and socialSkillMembers

Clean up any transcript so it reads like an article

Turn a messy auto-transcript into clean, quotable text with headings, without changing what was said.

Version 1.2Updated 12 Oct 2026

Needs access to
None
Permissions
None
Time to result (our estimate)
3 minutes to set up, 2 minutes a run

Fit

About this item

Auto-transcripts are full of "um", broken sentences and run-on paragraphs. This skill cleans them up so they read like an article: filler gone, punctuation fixed, headings added every few minutes, and the five most quotable lines pulled out at the end.

YouTube auto-captions work too, pasted straight from "Show transcript". They arrive with no punctuation, no capitals and a timestamp every few seconds; the skill turns them into sentences and marks any word it cannot make out instead of guessing.

It keeps your words. Nothing is added or reworded beyond what clean reading needs.

What you get

  • A clean transcript with headings
  • Speaker labels kept
  • Five quotable lines

Good fit

  • Turning videos, podcasts and calls into posts or notes

Not for

  • Legal or medical records that must stay word for word

Who it is for

Skill level: Use AI weekly

How to use

How to use it

In Claude

  1. Download the skill zip.

  2. In Claude on the web or the desktop app, open Settings, then Capabilities, and turn on Code execution and file creation. On a Team or Enterprise plan, an owner turns this on first.

  3. Open Customize, then Skills. Choose +, then Create skill, then Upload a skill, and pick the zip.

In Claude Code

  1. Download and unzip the skill.

  2. Move the transcript-cleanup folder into your Claude Code skills folder.

    macOS and Linux
    ~/.claude/skills/transcript-cleanup/
    Windows
    %USERPROFILE%\.claude\skills\transcript-cleanup\
  3. Start a new Claude Code session. It loads the skill when the task needs it.

In ChatGPT

  1. Download and unzip the skill.

  2. In ChatGPT, choose New project in the sidebar.

  3. Open Project settings from the project menu and paste SKILL.md from the transcript-cleanup folder into its instructions.

  4. Add the skill's other files (anything in its references or assets folder) to the project. Then chat inside the project.

Same steps in

  • Cursor
  • Codex
  • Gemini CLI
  • GitHub Copilot
  • VS Code
  • Amp
  • Goose
  • OpenCode
  • Junie
  • Roo Code
  • Kiro
  • Letta
  • Manus
  • OpenClaw
  • Hermes
  • Muse
  • Dots
  1. Download and unzip the skill.

  2. Move the transcript-cleanup folder into your skills folder.

    macOS and Linux
    ~/.agents/skills/transcript-cleanup/
    Windows
    %USERPROFILE%\.agents\skills\transcript-cleanup\
  3. Start a new session in your AI tool. It loads the skill when the task needs it.

In Grok

  1. Download the skill zip.

  2. Add the zip in the skills or project settings of Grok.

Trust

What's inside

  • transcript-cleanup folder
    • references folder
      • what-to-remove.md807 B
    • SKILL.md2.8 KB

File names and sizes only. The files come with membership.

Permissions required

None

Requirements

You bring

  • A transcript as text: YouTube auto-captions, or the .txt, .srt or .vtt file your recorder saves

Tested on

VerifiedTested in the last 90 days
  • Claude · Claude Opus 5.512 Oct 2026
  • Claude18 Sep 2026
  • Claude Code18 Sep 2026
  • ChatGPT15 Sep 2026

Known limitations

  • It works from text. It cannot listen to audio or watch video.
  • Very long recordings (over about two hours) are best split in two.

More

Used in these systems

Questions

Does it change what I said?
No. It removes filler words, false starts and repeats, and fixes punctuation. It never adds a point or changes a meaning. Unclear passages are marked, not guessed.
Which transcripts work?
YouTube auto-captions, plain text, .srt, .vtt and .sbv. For YouTube, open the video in a browser, click "...more" under the title, then "Show transcript", and copy the text. Timestamps are kept at each heading so you can find the moment in the video.
Where is it used in the library?
It is the first step of the video to posts playbook.