yt-script

---
name: yt-script
description: "Develop YouTube scripts and visual opening hooks; use for new script or hook work, not approved-frame repair."
---

# yt-script

## Context and execution
Read the current project's instructions, channel profile and approved creative decisions when available. Match the user's language and channel audience. Do not require a published-video portfolio from a new creator. Keep existing canonical project documents current rather than creating redundant profiles.
Distinguish narrative film, spoken explanation and mixed formats. Never impose narration on visual storytelling or automatically remove meaningful silence. Preserve approved characters, story and shot continuity.
Use actual data for numbers, timestamps and performance claims. Label hypotheses and lexical scores. Python tools require Python 3.10 or newer; invoke them by absolute path or from the repository root. Keep all eleven sibling skill folders together because some tools share resources. If execution, browsing, media inspection or private analytics are unavailable, state the limitation and work from supplied evidence; never pretend to run a tool.
These workflows draft and analyze. They do not publish videos or send replies. External posting requires explicit user authorization.

## Workflow

Identify narrative film versus spoken explainer first. For a film, derive action, stakes, question and causal payoff from the approved premise. Draft several materially different visual openings; compare immediate readability, curiosity, emotional stakes and promise-to-payoff fit in words, not a fabricated success score. Preserve the user's requested number of options. Carry the chosen opening into timed story beats with visible action, meaningful silence, optional motivated dialogue and a complete ending. Estimate duration from blocking and shot timing, not speech rate alone.
For English spoken hooks only, [hooks.json](hooks.json) supplies 21 optional patterns. Examples are illustrative, not verified factual claims. Run `python <skill>/hookscore.py --hook "..." --json` as a lexical diagnostic, never as a virality predictor or a visual-hook score. A low address score does not justify adding direct audience address to fiction. For spoken material estimate reading time separately from pauses/actions. Deliver the script and the reason for the selected opening.

• SKILL.md 2 KB
• hooks.json 10 KB
• hookscore.py 6 KB
• agents/openai.yaml 1 KB