End-to-end YouTube script writing skill for Claude Code, encoding the Iman Gadzhi / Educate.io scripting process documented by Kyle (product team lead at Consulting / Educate.io).
This skill turns Claude into a structured YouTube scriptwriter that walks an operator through the exact same workflow Iman's team uses internally — brief → narratives → outline → context-loaded chat → section-by-section writing → quality gate → trim.
When a user asks Claude (running in Claude Code) to "write a YouTube script", "draft an intro for my video", "script the body points", or any similar request — Claude loads this skill and follows the encoded 10-step workflow:
- Preflight — title, channel, brief, campaign objective, narratives, mechanism+contributor.
- Receive brief — title only / title + bullets / title + meeting transcript.
- Confirm narratives — attention / validation / democratization / timing (channel-modified).
- Pick length & depth (never told to Claude — manual trim later).
- Build outline manually — Intro → Packaging → Body Points → Outro, chronological bullets.
- Open fresh Claude chat + paste context prompt — 3 blocks + the load-bearing "two-three layers deeper / wait for further instruction" closer.
- Load reference material — high-performing transcripts on the same topic + creator's natural-speech transcripts + optional research.
- Write section by section — Intro (One-Sentence Persuasion OR Hook→Break Belief→X-to-X→Credibility) → Packaging (open/close loops) → Body Points (Value Loop: Context → Application → Framing) → Outro. Paste prior approved sections each time.
- Quality control gate — 7-question checklist + banned-language sweep.
- Manual length trim (never ask Claude to shorten).
youtube-scripting/
├── SKILL.md # main runbook (Claude reads this when activating)
├── README.md # this file
└── reference/
├── context-prompt.md # exact opening prompt template
├── narratives-and-channels.md # 3 channels + 4 narratives explained
├── copywriting-structures.md # both intro structures + packaging loops + Value Loop
├── word-tracks.md # exact correction prompts for fixing punchy AI output
├── banned-ai-language.md # comprehensive banned-words sweep list
└── quality-checklist.md # 7-question gate before finalising
This skill installs into Claude Code's user-skills directory at ~/.claude/skills/. Once installed, Claude Code automatically loads it whenever a relevant prompt comes in (the description in SKILL.md frontmatter is what controls activation).
# Make sure ~/.claude/skills/ exists
mkdir -p ~/.claude/skills
# Clone this repo into the skills folder
cd ~/.claude/skills
git clone https://github.com/laurentickner/youtube-scripting-skill.git youtube-scriptingThat's it. Claude Code picks up the skill on the next prompt.
cd ~/.claude/skills/youtube-scripting
git pullIn Claude Code, run:
/skill youtube-scripting
…or just say "write a YouTube script for [title]" and Claude will activate it automatically.
Send your teammate this repo URL: https://github.com/laurentickner/youtube-scripting-skill
They run the install command above. The skill works the same way for them as it does for you — no per-user config, no API keys, no setup beyond git clone.
If they want updates as you improve the skill, they re-run git pull whenever they want.
- Model: Claude Sonnet 4.6 Extended Thinking (the team's tested baseline). Opus only for unusually high-stakes campaigns.
- One script = one chat. Never start a new chat partway through. The compounding learning across the conversation is what makes later sections cleaner.
- Always paste prior approved sections when requesting the next one.
This skill is a faithful encoding of the YouTube scripting process documented by Kyle (product team, Consulting / Educate.io) in the Loom + Notion page covering:
- The 7-part scripting process
- The "two/three layers deeper" Claude-laziness fix
- The peeling-onion correction word-track
- The anti-AI-jargon correction word-track
- The Value Loop body-point framework (originally from creator Callaway)
- The One-Sentence Persuasion intro structure
- The Hook → Break Belief → X-to-X → Credibility intro structure
- The opening / closing retention loops in packaging
- The meet-AND-exceed expectations rule for intros
- The full banned-AI-language list
- The 7-question quality control checklist
- Channel personas (Main / Business / Extended) and how they modify narrative intensity
- The 4 narratives (attention / validation / democratization / timing)
- Unique mechanism + contributor architecture
Credit to Kyle and the Educate.io product team for the underlying playbook.
The skill is versioned in git. To propose an update:
- Edit the relevant
.mdfile in~/.claude/skills/youtube-scripting/. git add . && git commit -m "your update"git push
If your teammate is also using the skill, ping them to git pull.