create-viral-content

An Agent Skill that turns a draft or a brief into social content that spreads
and reads as written by a person with a point of view. It covers hooks,
platform rules, titles and thumbnails, AI-tell removal, adversarial
refinement and AI-disclosure rules, updated for the platform changes of
June to November 2026.

Install

The skill is a folder named create-viral-content containing SKILL.md and
references/. Every platform below needs the folder name to stay exactly
create-viral-content.

claude.ai (web and desktop). Build a zip whose single top-level folder
is create-viral-content, then upload it under Settings, Capabilities,
Skills:

git clone https://github.com/aaaronmiller/create-viral-content
cd create-viral-content && git archive --prefix=create-viral-content/ -o ../create-viral-content.zip HEAD

Do not upload GitHub's "Download ZIP" archive: it wraps the files in
create-viral-content-main/, and the folder name must match the skill name.

Claude Code.

git clone https://github.com/aaaronmiller/create-viral-content ~/.claude/skills/create-viral-content

Codex, Gemini CLI, opencode, Qwen Code, Cursor, Kilo (shared user-level
skills folder):

git clone https://github.com/aaaronmiller/create-viral-content ~/.agents/skills/create-viral-content

Antigravity: clone into ~/.gemini/config/skills/create-viral-content.

Claude API and OpenAI API: upload the same zip as a skill bundle.

Use

Write a LinkedIn post about [topic]. I ran [experiment] for [time] and
[result]. Audience: [who].

Give me 10 YouTube title and thumbnail-text variants for a video about
[topic]. The video shows [what it delivers].

The skill runs an eight-step workflow: pin the brief and the author's own
stake, check platform rules, draft hooks, draft the body, remove AI tells,
run five refinement passes, package variants, and check disclosure.

Files

File	Purpose
`SKILL.md`	Workflow, platform rules, hooks, titles, top AI tells, ethics and disclosure
`references/platform-updates-2026.md`	Dated 2026 ranking, anti-AI and disclosure changes with sources
`references/ai-tells.md`	100+ AI tells, including current-model tells (2026)
`references/platform-templates.md`	Templates for Reddit, X, LinkedIn, YouTube, TikTok, Instagram, email
`references/refinement-protocol.md`	Adversarial passes for text and thumbnails
`references/viral-titles.md`, `references/title-formulas.md`	Title psychology and 50+ formulas
`references/viral-thumbnails.md`, `references/thumbnail-checklist.md`	Thumbnail design and checklist
`references/research-statistics.md`	Research data, with 2026 studies that supersede older rules
`references/humanize-integration.md`	Hand-off to the humanize-writing skill
`references/cross-platform-adaptation.md`	One piece across several platforms
`references/trend-intelligence.md`	Finding a topic when you have none
`references/feedback-loop.md`	Tracking results over time
`CHANGELOG.md`	Version history
`marketplace.json`	Marketplace manifest

Optional companion

humanize-writing
adds a voice-polish pass. Without it, the skill uses its built-in checklist.

Ethics

Use it to sharpen your own ideas. It refuses astroturfing, misinformation,
impersonation, fake reviews and invented personal experiences, and it tells
you which AI-disclosure rule applies (including EU AI Act Article 50 for
public-interest content from 2026-08-02).

Version

Field	Value
Version	3.0.0
Updated	2026-11-10
Author	ice-ninja
License	MIT
Format	Agent Skills (SKILL.md with name, description, license, metadata)

Issues and suggestions: open an issue on this repository.