Ripple

A Claude Code skill that runs every morning, fetches what's trending, scores trends against your niches, and drops ready-to-post copy for Instagram, LinkedIn, TikTok, and X into Notion.

Built on the same architecture as trendvane — same trend pipeline, different output: real platform-native post copy instead of video briefs.

• Instagram — full caption with hook, body, CTA + 20–30 hashtags for first comment
• LinkedIn — professional post, value-first opening, discussion CTA, 3–5 hashtags
• TikTok — text overlay (≤10 words), display caption, full caption, audio direction, hashtags
• X — single tweet (≤280 chars) or numbered thread (2–7 tweets)
• Trend Intel — what it is, why it scored, source URLs

All 12 posts are voice-audited against your examples before pushing to Notion.

# Recommended
npx skills add your-username/ripple -g -a claude-code

# Manual
git clone https://github.com/your-username/ripple
cd ripple
python3 setup.py

Then open Claude Code and type /ripple for the guided setup.

When you run /ripple for the first time, Claude walks you through:

1. Notion — creates your Ripple database and collects workspace/database IDs
2. Niches — you define 2–5 niches with keywords, weights, and optional subreddits
3. Voice profile — you paste 2–3 example posts per platform; Claude matches your voice
4. API keys — optional: YouTube, Reddit, Product Hunt (Ripple works without them)

After setup, run /ripple any time.

Each trend is scored 0–100 across 5 dimensions:

Weights are overridable in my_niches.json. At least 1 of the 3 daily picks always comes from your highest-weighted niche.

All user config lives in ~/.claude/ripple/ and is never overwritten by updates.

Your Ripple database needs these properties:

Each page body contains: IG → LinkedIn → TikTok → X → Trend Intel.

On demand:

/ripple

Or: "what should I post today?" inside Claude Code.

Scheduled (daily):

# 5:30am fetch + 6:00am generate
# Set up via cron or the Claude Code /schedule skill

Score only (skip fetch):

python3 ~/.claude/skills/ripple/scripts/run_all.py --score

Before pushing to Notion, Claude re-reads all 12 posts and checks:

• Zero AI-tell phrases (delve, leverage, game-changing, etc.)
• Matches your voice_examples.md patterns
• Per-platform format rules (char limits, hook placement, hashtag counts)
• No brand-unsafe content (politics, alcohol, gambling, explicit, competitor attacks)

• Claude Code (npm install -g @anthropic-ai/claude-code or claude.ai/code)
• Notion MCP configured in Claude Code
• Python 3.10+
• Notion workspace with a Ripple database

ripple/
├── .claude-plugin/plugin.json
├── .github/workflows/ci.yml
├── skills/ripple/
│   ├── SKILL.md                    # Full workflow — what Claude reads
│   ├── requirements.txt
│   ├── references/setup_guide.md   # Step-by-step first-run flow
│   ├── scripts/
│   │   ├── common.py               # Shared utils, TrendCandidate class
│   │   ├── run_all.py              # Orchestrator: fetch + score
│   │   ├── score_trends.py         # Scoring + clustering engine
│   │   ├── validate_setup.py       # Pre-flight config check
│   │   ├── fetch_hackernews.py
│   │   ├── fetch_reddit.py
│   │   ├── fetch_google_trends.py
│   │   ├── fetch_youtube.py
│   │   ├── fetch_producthunt.py
│   │   ├── fetch_tiktok.py
│   │   └── push_to_notion.py       # Notion schema reference
│   ├── templates/
│   │   ├── post_template.md        # Output format for all 4 platforms
│   │   └── platform_rules.md       # Per-platform constraints
│   ├── memory/                     # Shipped defaults (read-only)
│   │   ├── my_niches.json
│   │   ├── voice_examples.md
│   │   ├── notion_config.example.json
│   │   ├── secrets.example.json
│   │   └── used_angles.json
│   └── tests/test_scoring.py
├── README.md
├── setup.py
├── pyproject.toml
└── requirements.txt

# Lint
ruff check skills/ripple/scripts skills/ripple/tests

# Fix
ruff check --fix skills/ripple/scripts

# Test
python3 -m unittest discover -s skills/ripple/tests -v

MIT