AI-powered YouTube comment analysis for brands, marketers, and creators. Analyze comments, detect threats, generate smart replies, and measure KOL collaboration ROI — all from your terminal with Claude Code.
Built by the team behind ReplyCue AI — the AI comment moderation platform for YouTube.
| Skill | Description | Use Case |
|---|---|---|
| YouTube Comment Analyzer | Analyze video comments with 5-layer brand relevance scoring | "Is anyone talking about my brand in this KOL video?" |
| Smart Reply Generator | Generate brand-appropriate replies in 3 tones | "Help me write a professional reply to this complaint" |
| KOL ROI Report | Comprehensive collaboration effectiveness report | "Was this $5K sponsorship worth it?" |
| Brand Keyword Extractor | Extract keywords from any website for SEO and monitoring | "What keywords should I track for my brand?" |
| Comment Threat Scanner | Detect scams, fake claims, and brand safety risks | "Are there any dangerous comments I need to handle?" |
You need two API keys:
# Required for all skills
export ANTHROPIC_API_KEY="sk-ant-..."
# Required for YouTube-related skills (1, 3, 5)
export YOUTUBE_API_KEY="AIza..."- Get your Anthropic API key at console.anthropic.com
- Get your YouTube API key at Google Cloud Console (enable YouTube Data API v3)
Clone this repo and add the skills directory to your Claude Code configuration:
git clone https://github.com/YOUR_USERNAME/replycue-ai-skills.git
# Add to your project's .claude/settings.json
# or copy skills/ into your .claude/skills/ directoryJust describe what you want in natural language:
> Analyze the comments on this YouTube video for my brand ShieldVPN:
https://www.youtube.com/watch?v=VIDEO_ID
> Generate a professional reply to this comment:
"Your product crashed 3 times today, worst purchase ever"
> Give me a KOL ROI report for this sponsored video:
https://www.youtube.com/watch?v=VIDEO_ID
Brand: Notion, Collaboration type: sponsored
> Extract brand keywords from https://notion.so
> Scan this video for brand safety threats:
https://www.youtube.com/watch?v=VIDEO_ID
Brand: Notion
Our 5-layer brand relevance model goes beyond simple keyword matching:
Final Score = 50% Semantic Score ← Does the comment text relate to the brand?
+ 15% Video Relevance ← Is the video itself about the brand?
+ 15% Entity Match ← Is the brand name explicitly mentioned?
+ 10% Intent Score ← Is this a purchase intent, question, or risk?
+ 10% Context Score ← How deep is the brand integration in the video?
Different collaboration types have different relevance thresholds:
| Type | Threshold | Why |
|---|---|---|
| Dedicated (full review) | 50% | Most comments on a dedicated video are relevant |
| Sponsored (segment) | 60% | Filter off-topic discussion |
| Mention (brief) | 85% | Only clearly brand-related comments pass |
| Placement (background) | 85% | Product is barely discussed |
Read the full scoring model documentation →
All analysis skills support multiple output formats:
- Markdown (default) — Human-readable reports
- CSV — Import into spreadsheets or databases
- JSON — Integrate into your own tools and pipelines
All skills support comments in any language:
- English, Chinese (中文), Japanese (日本語), Korean (한국어)
- Hindi (हिन्दी), Arabic (العربية), Spanish, Portuguese
- And 20+ more languages
Output language automatically matches your preference or the comment's language.
The flagship skill. Fetches comments from any YouTube video and classifies each one into 7 business categories:
| Category | What it means | Priority |
|---|---|---|
purchase_intent |
User wants to buy | Medium |
product_question |
Questions about features, pricing, bugs | Medium |
brand_risk |
False claims, scams, privacy concerns | High |
competitor_mention |
Compares with competitors | Medium |
positive_feedback |
Praises the product | Low |
negative_feedback |
Complaints, disappointment | Medium |
irrelevant |
Not about the brand | — (filtered) |
Generates 3 reply options per comment with strategy explanations:
- Professional — Clear, direct, respectful
- Friendly — Warm, approachable, personable
- Casual — Conversational, like texting a friend
Full collaboration analysis including:
- Engagement quality score
- Purchase signal strength
- Brand risk level
- Audience-brand alignment
- Actionable recommendations
Extracts 5 categories of keywords:
- Primary keywords (high-value, core brand terms)
- Secondary keywords (related, broader reach)
- Long-tail phrases (with search intent classification)
- Competitor keywords
- Negative keywords (terms to exclude)
Focused brand safety scan with 3 severity tiers:
- HIGH — Scams, impersonation, false claims, doxxing
- MEDIUM — KOL misrepresentation, privacy concerns, competitor attacks
- LOW — Negative experiences, general toxicity, spam
All skills use claude-haiku-4-5-20251001 for cost efficiency:
| Skill | Typical API Cost | Comments Analyzed |
|---|---|---|
| Comment Analyzer | ~$0.02-0.05 | 100 comments |
| Smart Reply | ~$0.001 | 1 comment |
| KOL ROI Report | ~$0.03-0.08 | 100 comments + summary |
| Keyword Extractor | ~$0.002 | 1 website |
| Threat Scanner | ~$0.02-0.05 | 100 comments |
These skills are great for one-time analysis. For ongoing, automated YouTube comment monitoring:
ReplyCue AI provides:
- Real-time comment monitoring across multiple videos
- Automatic AI analysis as new comments arrive
- Team collaboration with role-based access
- One-click AI reply generation
- Dashboard with business intelligence
- Alert rules for high-priority comments
- Multi-video KOL performance comparison
MIT License — use these skills freely in your projects.
Found a bug or have an improvement? PRs welcome!
- Fork this repo
- Create your feature branch (
git checkout -b feature/improvement) - Commit your changes
- Push and open a PR