Viral Titles for Agents: A Practical Skill Worth Testing

Where this fits

Your agent just got asked: “Give me 20 titles for my next YouTube video.” It returns a generic list. That is not enough if the same content also needs titles for WeChat, X, and Bilibili. viral-title gives agents a repeatable workflow: generate, score, select, and adapt title candidates per platform.

Why agents benefit

• Structured generation: Phase 1 produces 30 candidates from 10 universal formulas x 3 variants, so the agent explores diverse title patterns instead of guessing.
• Built-in scoring: candidates are scored automatically, and the agent can recommend one best title with a clear rationale.
• Platform methods: dedicated handling for 公众号, X/Twitter, YouTube title-thumbnail packaging, and Bilibili title-cover-tag packaging.
• Title-library reuse: previously tested titles can be stored and reused for series, avoiding cold starts.
• Evolution loop: lightweight feedback lets the workflow improve future title suggestions based on what performed.

Practical scenarios

Batch titles for a content calendar

A media team asks for 20 headline options for each of five WeChat articles. The skill generates a full board using 10 formulas, scores each candidate, and leaves editors with a ranked shortlist instead of a blank page.

Cross-posting one video to four platforms

A creator has a single YouTube video and wants adapted titles for X, WeChat, and Bilibili. viral-title packages hooks, thumbnails, and tags where relevant, rather than just translating the original title.

Reusing proven title structures

A weekly newsletter agent uses the title library to keep a consistent voice. When click-through data comes back, the evolution loop refines the formulas for the next issue.

Add it to your agent workflow

Install the skill with:

npx skills add kangarooking/kangarooking-skills --skill viral-title

Then use it in an agentic writing step:

User: I have a video about running local LLMs. Target: Python developers. Platform: YouTube.
Agent: Generates 30 candidates, scores them, and returns:
1. "Run 70B LLMs on a Laptop Without a $10K GPU" (score 8.7)
2. "Local LLMs Are Finally Practical for Python Devs" (score 8.2)
...

If the request is vague, the skill asks one clarifying question before generating.

Compare before adopting

Compare viral-title with:

• Quality signals: run a few A/B tests and see whether scored titles outperform your old prompt-generated ones.
• Maintenance freshness: title formulas decay; check how often the repository is updated.
• Alternatives: a plain LLM prompt may suffice for low-volume work; platform analytics can cover selection better.
• Workflow fit: title-library reuse only helps if your agent has persistent memory and a publishing pipeline to write back to.

Why it is worth tracking

viral-title is worth tracking because it fixes a concrete gap: most agents can write a title, but few can systematically generate, score, and adapt titles across platforms. With 568 stars on GitHub, it has more community validation than a custom prompt. It is especially relevant if you build agents for WeChat, Bilibili, or cross-platform content operations. The skill is listed in the OpenAgentSkill registry at https://www.openagentskill.com/skills/kangarooking-viral-title, so you can compare it with similar skills. Re-evaluate after a few production runs to confirm the generated titles actually improve engagement.