YouTube Metadata That Actually Converts: The youtube-copy Skill

Where this fits

You've just finished editing a 15-minute tutorial, but the video is nothing without the metadata that gets it discovered. You need a title that earns the click, a description that ranks, tags that categorize, and chapters that keep viewers watching. Manually crafting all four from a transcript is tedious and inconsistent. The youtube-copy skill automates this: feed it a timestamped SRT or even just a topic, and it outputs a structured youtube.md ready to paste into YouTube Studio.

Why agents benefit

• Grounds metadata in source truth – The skill never invents claims; it extracts only what's supported by the transcript or provided context, reducing hallucination risk.
• Handles multiple input types – Whether you pass a full SRT, a PR description, or a freeform topic, the agent can adapt and still produce all four metadata pieces.
• Enforces timestamped chapters – It knows chapters require timestamps and will ask for them or skip gracefully, avoiding broken output.
• Strict formatting compliance – The skill outputs a consistent youtube.md structure, making it easy to pipe into publishing pipelines or other tools.
• No em-dash policy – A quirky but practical constraint that keeps copy clean for YouTube's character limits and style guidelines.

Practical scenarios

1. Post-publish optimization for a tech channel

You have a recording of a coding walkthrough. The agent receives the SRT and a short brief. It returns a title that includes the main keyword, a description with timestamps and links, relevant tags, and chapters. You paste it in and see a lift in CTR within days.

2. Repurposing a blog post into video metadata

You maintain a blog and want to turn a popular article into a YouTube video. Instead of writing metadata from scratch, you feed the article's text (as a freeform description) to the skill. It generates a title and description that align with the article's SEO, saving you an hour of manual work.

3. Batch processing for a content agency

You manage multiple client videos. Each one has a transcript but no metadata. Your agent loops through them, invoking youtube-copy per video, and writes a youtube.md into each client's output folder. Consistency is guaranteed across all deliverables.

Add it to your agent workflow

Install the skill via npx:

npx skills add AlemTuzlak/skills --skill youtube-copy

Then trigger it in your agent's instructions with a prompt like:

Use youtube-copy on this SRT: [paste transcript] – write youtube.md in the output folder.

The skill will generate the four metadata sections and save them to disk. For a deeper look at how this skill fits into a broader agent toolkit, check out the skill's page on OpenAgentSkill.

Compare before adopting

Before relying on this skill, compare it against a few baselines:

• Quality signals – Test on your own videos: does the title match your brand voice? Are chapters accurately aligned with timestamps?
• Maintenance freshness – Check the GitHub repo for recent commits. A skill that hasn't been updated in months may lag behind YouTube's algorithm changes.
• Alternatives – Could you use a generic LLM prompt instead? This skill adds structure and guardrails, but if your needs are trivial, a simple prompt might suffice.
• Workflow fit – Does it integrate cleanly with your existing agent's output handling? The strict youtube.md format is a plus if you already use file-based outputs.

Why it is worth tracking

With 39 stars and an active author, youtube-copy is a niche but growing skill. Its focus on timestamped chapters and source-grounded copy makes it a solid choice for developers building content-publishing agents. Watch how it evolves: if it adds support for multiple languages or A/B title testing, it becomes even more valuable. Evaluate it when you need reliable, structured YouTube metadata without reinventing the wheel.