Youtube Packaging

YouTube Packaging

Input: a topic plus any available footage, assets, audience, and proven data. Output: 5 title + thumbnail pairs, one image prompt per pair, and one recommended pick. Use these skill-specific defaults for package output; apply broader style heuristics only when they fit the actual video.

Steps

1. State the viewer promise in one sentence. Note what the planned video can actually show.
2. Draft 5 concise, front-loaded titles. Use varied angles supported by the footage; numbers, fear, and curiosity are optional, never filler. Make one of the 5 fit the Galloway formula: subject + quantified result + timeframe + mechanism promise ("How Beluga Gained 4M Subs in 3 Months (Genius Strategy)").
3. Sketch distinct concepts for the same promise (for example, real proof, the workflow, or a consequence) before refining crops, colors, or text. Each thumbnail should add useful information; it may repeat a necessary name or unit.
4. Apply the truth, legibility, feed-context, and comprehension checks in [Visual checks and example](VISUAL-CHECKS.md). Truth and readable meaning are hard gates.
5. Compare passing pairs on appeal, clarity, distinctiveness, and brand fit. These are subjective editorial judgments, not CTR predictions. Recommend the strongest fit; explain the choice briefly.
6. Name the backup. Title and thumbnail are a critical component (MrBeast handbook): the pick ships, the runner-up is the swap if the first-day CTR disappoints. Upload up to 3 thumbnails to YouTube Test & Compare; it picks the winner on watch-time share, not clicks.

When metrics are available, do not diagnose packaging from a universal CTR cutoff. Read CTR alongside traffic source, impressions, audience, timing, and comparable channel history.

Title and thumbnail defaults

• Keep titles succinct and put the main promise early. 45–55 characters is a drafting preference, not a minimum; do not pad a clear title.
• Prefer plain words and active voice. Avoid "Nobody Tells You," "Ultimate Guide," and exclamation marks.
• Use at most 3 main visual elements. Text is optional; default to 4 words or fewer, with a little more allowed when it stays plainly readable.
• A face is optional. Choose a supplied, real photo only when the person's identity or expression helps; an object-led image can carry the story.
• During filming, capture useful real frames of the face, hardware, and result when available. For image prompts, specify subject/object, any exact text, background, framing, and the real asset to use. Never invent the creator's identity, footage, data, screenshots, benchmarks, or results. Conceptual illustrations and fictional people are allowed when clearly illustrative and not presented as captured evidence.

Output format

Use a table with title, character count, thumbnail concept, image text, truth/readability gate, and brief editorial notes. Then give the pick and one image prompt per pair. Never recommend a pair that fails a hard gate.

Changelog

• 2026-09-20: Added evidence, concept, comprehension, and feed-context gates; made style rules contextual and editorial scores explicitly subjective.
• 2026-09-21: Added the Galloway title formula slot, the backup pair rule, Test & Compare, and the vidIQ 2026 breakout medians as a checklist in VISUAL-CHECKS.md.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: ravsau
• Source: ravsau/youtuber-skills
• License: MIT

Install and usage instructions live in the source repository linked above.