Get started
Short-Form Script Gates

Short-Form Script Gates

Pre-publish quality gates for Shorts, Reels, and TikTok scripts. Rejects hooks whose first line can't stand alone (conditional clauses, bare noun phrases), scripts with no reason to watch today (no share trigger), broken structure (not hook → 3 facts → twist → close), and subtitle violations that die on a muted phone screen. Distilled from running a daily short-form knowledge channel. Every REJECT comes with a concrete rewrite, not just a verdict.
#Vídeo#Redes sociais#Escrita
Avaliação
São necessárias mais avaliações
Vendas
0
Como usar
Baixar

Short-Form Script Gates

Pre-publish quality gates for short-form video scripts (YouTube Shorts, Reels, TikTok), distilled from running a daily-output faceless knowledge channel. Each gate encodes a measured failure: episodes that violated these patterns died in the feed; episodes that satisfied them were the outliers that traveled.

What it does

Install this skill and your agent reviews (or writes) every script against five hard gates:

Gate Catches
1. First line stands alone Feed viewers decide in ~1 second, reading only line one. Conditional clauses ("If you wash your bedding weekly...") and bare noun phrases ("The truth about airplane coffee") defer the payoff to a line nobody reads.
2. A reason to watch TODAY The hook or twist must contain same-day relevance, a deadline, or an action directive — as a full sentence. The strongest predictor of sharing: people forward videos as a favor.
3. Structure Hook → exactly 3 facts → twist (after ≥2 facts) → closing identity line. Five facts is padding; twist-first is just a claim; ending on the twist drops your sign-off. One video, one topic.
4. Works muted, works blind Subtitles + visuals alone must carry the story, and narration + visuals alone must too. Wrong examples shown on screen get an explicit ✘ overlay — a muted scroller who catches one frame remembers the error as fact.
5. Subtitle discipline Subtitles = narration's exact words; max 1 highlight per line; no emoji/fraction glyphs that break rendering; "=" only for numeric conversions; no symbols in narration (TTS skips them silently).

Every REJECT comes with a concrete rewrite, not just a verdict.

Example

Input:

Review this Shorts script — why isn't it performing?

HOOK line 1: "If you wash your bedding every single week..."
line 2: "dust mites might still be alive"
[5 facts, ends on the twist, subtitles with "=" and a ⅔ glyph]

Output (abridged):

## Script Gate Report

Gate 1 (first line): REJECT — conditional clause, payoff deferred to
  line 2. Rewrite: "Dust mites survive your washing machine." /
  line 2: "Even if you wash every single week."
Gate 2 (share trigger): REJECT — no today/deadline/action sentence.
  Honest fix: "Before your next wash, check your machine's default
  temperature — it's probably 30-40°C."
Gate 3 (structure): REJECT — 5 facts (cut 2 for a follow-up video);
  ends on the twist with no closing line.
Gate 5 (subtitles): REJECT — "=" used twice and non-numerically;
  ⅔ glyph breaks rendering; 3 highlighted words on one line.

Verdict: FIX FIRST — rewrite skeleton included.

Who it's for

  • Short-form creators and editors (Shorts / Reels / TikTok)
  • Social video teams
  • Anyone whose scripts are "fine" but don't survive the first second — or never get shared

What you get

A single SKILL.md — clear, auditable instructions your agent follows. No scripts, no network calls, no credentials, no data leaves your machine. Works with Claude Code and SKILL.md-compatible agents.

FAQ

Will it write scripts too, or only review them?
Both. When writing from scratch, it satisfies the gates during drafting and appends an honest self-review — if a gate can't pass for your topic, it says so and explains the trade-off.

Does it fabricate urgency to force Gate 2?
No — fabricated deadlines are rejected as a fact-check failure. If a topic genuinely has no same-day angle, the skill flags it as feed-proof but share-dead and lets you decide.

Note: This skill is a set of instructions for an AI agent; output quality depends on the underlying model. Gates encode one channel's measured experience — treat them as strong defaults, not laws of physics.