skills/create-viral-content/SKILL.md
AUTOMATIC ACTIVATION: Use this skill whenever generating content for social media platforms including Reddit posts, Twitter/X threads, LinkedIn posts, YouTube videos/comments, TikTok scripts, Instagram posts/reels, email subject lines, blog titles, or any content intended for public engagement. Also triggers on: "make this viral", "social media post", "catchy headline", "hook", "engagement", "shareable", "go viral", "attention-grabbing", "clickable", "scroll-stopping", "title ideas", "hot take", "more engaging", "subreddit", "thread". Apply to any AI-generated content facing public audience or hostile perception.
npx skillsauth add aaaaqwq/claude-code-skills create-viral-contentInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
⚠️ BEFORE USING THIS SKILL: Review all files in the
resources/directory. These contain AI tell catalogs, platform templates, refinement protocols, and 40-source research basis required for proper skill execution.
This skill synthesizes findings from 40 documented research sources:
Full statistics in resources/research-statistics.md.
Make your posts spread. This skill turns forgettable drafts into content that gets shares, comments, and action.
Good content doesn't come from one pass. You attack it, fix it, attack again:
Pattern: Prediction + Stakes
"I think [CONCEPT] is the [YEAR] [CATEGORY] that [OUTCOME]."
Example: "I think deliberative refinement is the 2026 prompt technique that matters most."
Why it works:
Pattern: Tribal Identity Split
"[TECHNIQUE] separates [WINNERS] from [EVERYONE ELSE]."
Example: "This separates serious builders from prompt tourists."
Why it works:
Pattern: Before/After Compression
"What used to require [OLD COMPLEXITY] now [NEW SIMPLICITY]."
Example: "What used to need 12 models chained together now takes one."
Why it works:
Required elements (in order):
Anti-patterns to avoid:
Pattern: Command, Not Request
BAD: "Try it. Change my mind." (beggy, engagement bait)
GOOD: "Your next [ACTION] shouldn't [OLD WAY]. It should [NEW WAY]."
Example: "Your next prompt shouldn't ask for an answer. It should demand: 'Attack this from three expert perspectives, ground your claims, then revise.'"
Why the command works:
Kill these on sight:
Title patterns that work:
Body guidelines:
Subreddit calibration:
Constraints: ~500 chars, must hook in first line, no formatting Pattern:
[Bold claim in first sentence]. [Mechanic in 2 sentences]. [Why now]. [Call to action or quotable closer].
Thread structure:
Per-tweet rules:
Patterns that perform:
Avoid:
Hook constraints: 1-3 seconds to capture, sound-off viewing common
High-performing hooks:
Caption optimization:
Constraints: Vertical 9:16, auto-play, first frame critical
Thumbnail (cover) matters less than:
Carousel posts:
Optimal length: 30-50 characters (mobile-first)
High-performing patterns:
Anti-patterns:
Done with structure? Run humanize-writing to polish the voice. Same viral hooks, human delivery.
If you've got both skills, call humanize-writing directly:
Apply the humanize-writing skill to this draft. Focus on:
- Removing AI vocabulary tells from the content
- Ensuring natural sentence rhythm
- Maintaining the viral hooks I've established
No humanize-writing? Run this instead:
| Platform | Humanization Level | Formality Target | |----------|-------------------|------------------| | Reddit | High | Casual expert | | LinkedIn | Medium | Professional but warm | | Twitter/X | Medium-High | Punchy, fragmentary OK | | YouTube | High | Accessible, conversational | | Hacker News | Medium | Technical, understated |
Check these numbers after humanizing:
resources/ai-tells.md)Match formality to platform:
Confidence calibration:
⚠️ CRITICAL: Titles determine 70% of content performance. Consult
resources/viral-titles.mdandresources/title-formulas.mdfor 50+ formulas.
Curiosity-Gap: "What [group] won't tell you about [topic]" Contrarian: "[Common belief] is dead. Here's what's next." Listicle: "[Number] ways to [achieve X] without [sacrifice]" How-To: "How to [achieve X] in [timeframe] (step-by-step)" Prediction: "[Concept] is the [year] [category] that [outcome]" Negative: "[Number] [topic] mistakes destroying your [metric]"
| Criteria | Score 0-3 | |----------|-----------| | Curiosity | "Must know" feeling? | | Specificity | Numbers, metrics? | | Emotion | High-arousal trigger? |
⚠️ CRITICAL: Thumbnails drive 70%+ of video performance. Consult
resources/viral-thumbnails.mdandresources/thumbnail-checklist.mdfor design protocols.
[person] with [shocked/surprised] expression, close-up portrait,
[vibrant color] background, studio lighting, high contrast,
YouTube thumbnail style, clean composition, no text
Before you ship, attack the draft:
Pass 1: The Skeptic "Why should I care? What's actually new here?"
Pass 2: The Expert
"Is this technically accurate? What would an expert nitpick?"
Pass 3: The Scroller "Would I stop scrolling for this? What's the hook?"
Pass 4: The Competitor "How is this different from the 10 similar posts?"
Pass 5: The Editor "What can I cut without losing meaning?"
Before (AI-generated feel):
I'm excited to share a revolutionary new productivity hack that will
change your workflow forever. Here's the thing: most people waste hours
on email. Let's dive into how inbox zero can transform your day. First,
you batch process. Second, you use templates. Finally, you schedule
check-ins. What do you think?
After (human voice):
Email before noon is self-sabotage. Tested this for 3 weeks. No inbox
until 2pm. My deep work hours went from 2 to 4+. That 7:47am Slack
ping? Not your fire. Morning brain builds. Afternoon brain reacts.
Flip the order and you're always playing defense. Two inbox windows:
2pm and 5pm. Handles everything that actually matters.
What changed:
testing
通用自媒体文章自动发布工具。支持百家号、搜狐号、知乎、微信公众号、小红书、抖音号六个平台的自动化发布流程。使用Playwright自动化实现平台导航和发布,支持通过storageState管理Cookie实现账号切换。
development
# SKILL.md - Model Configuration Status (mcstatus) ## 触发条件 - `/mcstatus` 命令 - 用户询问模型配备、模型配置、model status、模型列表等 ## 功能 实时生成 Agent + Cron 的模型配置报告,展示当前所有 agent 的主模型/fallback链和所有 cron 任务的模型分配。 ## 执行步骤 ### Step 1: 收集 Agent 模型配置 读取各 agent 的 models.json 获取主模型和 fallback 链: ```bash for agent in main ops code quant data research content market finance pm law product sales batch; do config=$(cat ~/.openclaw/agents/$agent/agent/models.json 2>/dev/null) if [ -n "$config" ]; then echo "=== $agent
tools
MCP 服务器智能管理助手。自动检测 MCP 可用性、智能开关、功能问答,提供人性化的 MCP 管理体验。
tools
从GitHub搜索并自动安装配置MCP(Model Context Protocol)服务器工具到Claude配置文件。当用户需要安装MCP工具时触发此技能。工作流程:搜索GitHub上的MCP项目 -> 提取npx配置 -> 添加到~/.claude.json -> 处理API密钥(如有)。