Self-evolving viral content trend advisor. Monitors 11 platforms, predicts what to post and when, and improves its own accuracy every week automatically.
Version: v0.6.4 | Languages: English / 中文
本 Skill 完整支持中文操作。安装完成后，你可以：
This skill fully supports both English and Chinese. During onboarding, the agent will ask which language you prefer.
ViralEvo monitors content platforms, scores trending topics using a weighted formula, predicts lifecycle windows, and automatically adjusts its own prediction weights every week based on how accurate it was.
Three core advantages over manual research:
After installation, add your Tavily API key:
echo "TAVILY_API_KEY=tvly-xxxx" >> ~/.openclaw/workspace/.env
Then tell your agent:
"Start ViralEvo setup"
— or in Chinese —
"开始趋势雷达设置"
Or run onboarding directly:
node {baseDir}/scripts/onboarding.js
When the user reports post results (e.g. "got 80k views", "效果很好"), the agent should:
python3 {baseDir}/scripts/feedback.py --search "<keyword>"python3 {baseDir}/scripts/feedback.py --topic-id <id> --platform <platform> --views <n>When the user says any of the following, the agent should run collect → report:
When the user says:
When the user describes post results, always match to a recent topic, confirm before logging:
Use the /trend feedback command or natural language — both are accepted.
| Command | Action |
|---|---|
node {baseDir}/scripts/onboarding.js | First-time setup wizard |
node {baseDir}/scripts/collect.js | Fetch trend signals from all sources |
python3 {baseDir}/scripts/report.py | Generate and output today's report |
python3 {baseDir}/scripts/verify.py --hours 24 | Verify yesterday's predictions |
python3 {baseDir}/scripts/verify.py --hours 72 | Verify 72h-old predictions |
python3 {baseDir}/scripts/weekly_review.py | Run self-evolution (Mondays recommended) |
python3 {baseDir}/scripts/keywords.py --show | View your keyword index |
python3 {baseDir}/scripts/keywords.py --add "term" | Add a keyword manually |
python3 {baseDir}/scripts/keywords.py --remove "term" | Remove a keyword |
python3 {baseDir}/setup.py | Check all system requirements |
python3 {baseDir}/scripts/feedback.py --list | List recent topics to log feedback for |
python3 {baseDir}/scripts/feedback.py --search "keyword" | Find a topic by keyword |
python3 {baseDir}/scripts/feedback.py --topic-id <id> --platform tiktok --views 80000 | Log post performance |
python3 {baseDir}/db/init_db.py | Re-initialize database (use if DB is corrupted) |
python3 {baseDir}/scripts/status.py | Quick health check — config, API key, DB, recent data |
| Requirement | Minimum | Role |
|---|---|---|
| Node.js | v18+ | Data collection, onboarding |
| Python | 3.10+ | Scoring, reports, self-evolution |
| OpenClaw | v2026.1+ | Agent runtime, scheduling |
| Tavily API Key | Free tier | Indirect platform search |
Tavily free tier = 1,000 calls/month. Single niche daily usage ≈ 60–120/month.
| Platform | Method | Confidence Cap |
|---|---|---|
| HackerNews | Official Algolia API | 1.00 |
| Dev.to | Official API | 1.00 |
| Product Hunt | RSS | 1.00 |
| JSON API (public) | 0.90 | |
| YouTube | Tavily search | 0.70 |
| Twitter / X | Tavily search | 0.70 |
| Tavily search | 0.70 | |
| Tavily search | 0.70 | |
| TikTok | Tavily search | 0.65 |
| Tavily search | 0.65 |
AI/Tech · E-commerce · Beauty/Skincare · Fitness/Health · Finance · Gaming · Fashion/Lifestyle · Education · Real Estate · Pets · Custom (11 niches)
Total Score =
(Platform Signal Strength) × W1 [default 0.25]
+ (Engagement Velocity) × W2 [default 0.25]
+ (Cross-Platform Spread) × W3 [default 0.20]
+ (Niche Relevance Score) × W4 [default 0.15]
+ (Goal Alignment Score) × W5 [default 0.15]
Constraints: W1+W2+W3+W4+W5 = 1.0 exactly
Each weight: floor=0.08, ceiling=0.45
Max change per weekly review: ±0.05 (±0.10 after algorithm change detection)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔥 ViralEvo | AI/Tech | 2026-03-09 08:15
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔴 ACT NOW (Score > 80)
1. OpenClaw Security Issue — 135k instances exposed
████████████████████ 93% | Confidence: 0.85
📅 Detected 14h ago | Source: hackernews
⏰ Estimated window: ~42h remaining
🎯 Post: TODAY
🟡 PREPARE (Score 60–80)
2. OpenAI Government Surveillance Controversy
████████████████░░░░ 78% | Confidence: 0.74
📅 Detected 6h ago | Source: dev.to
⏰ Estimated window: ~68h remaining
🎯 Post: Tomorrow morning
🟢 EVERGREEN (Score < 60)
3. MCP Protocol Enterprise Adoption
████████░░░░░░░░░░░░ 44% | Confidence: 0.79
📅 Steady growth — no spike
⏰ Relevant: 30d+
🎯 Post: Any time this week
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 Model Health
Accuracy : 58% (44 predictions)
Sources : 6/6 ✅
Tavily usage : 112 / 1,000 this month
Keyword index: 1,203 terms
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
ViralEvo runs automatically via OpenClaw's cron system. After onboarding, add these four jobs to your OpenClaw cron config.
How to add cron jobs in OpenClaw:
Tell your agent:
"Add a cron job to run ViralEvo daily at 8am"
Or add manually to ~/.openclaw/openclaw.json:
{
"cron": {
"jobs": [
{
"id": "viralevo-collect-report",
"schedule": "0 8 * * *",
"commands": [
"node ~/.openclaw/workspace/viralevo/scripts/collect.js",
"python3 ~/.openclaw/workspace/viralevo/scripts/report.py"
]
},
{
"id": "viralevo-verify-24h",
"schedule": "5 8 * * *",
"commands": ["python3 ~/.openclaw/workspace/viralevo/scripts/verify.py --hours 24"]
},
{
"id": "viralevo-verify-72h",
"schedule": "10 8 * * *",
"commands": ["python3 ~/.openclaw/workspace/viralevo/scripts/verify.py --hours 72"]
},
{
"id": "viralevo-weekly-review",
"schedule": "0 8 * * 1",
"commands": ["python3 ~/.openclaw/workspace/viralevo/scripts/weekly_review.py"]
}
]
}
}
See OpenClaw docs: https://docs.openclaw.ai/automation/cron-jobs
Instead of using .env, you can configure your Tavily key via ~/.openclaw/openclaw.json:
{
"skills": {
"entries": {
"viralevo": {
"enabled": true,
"apiKey": "tvly-your-key-here"
}
}
}
}
Daily verification (5 min and 65 min after your report time): re-fetches topics predicted 24h ago, compares predicted lifecycle vs actual activity, records error.
Weekly review (every Monday at your report time):
reports/YYYY-MM-DD_weekly.md| Period | Expected Accuracy |
|---|---|
| Week 1–2 | 30–40% (cold start) |
| Month 2 | 55–65% |
| Month 3+ | 65–75% |
| Month 6+ | 75%+ |
Accuracy = prediction within ±20% of actual topic lifecycle.
~/.openclaw/workspace/viralevo/
├── config.json ← niche, weights, schedule
├── user_profile.json ← onboarding answers, language
├── data/
│ ├── trends.db ← SQLite database
│ └── backups/ ← daily snapshots, 7-day retention
├── reports/ ← daily + weekly markdown reports
└── logs/
└── execution.log
All data is stored locally on your machine. The skill makes outbound network requests only to fetch public trend signals:
ViralEvo provides probabilistic estimates based on publicly available signals. It does not guarantee specific outcomes in views, impressions, followers, or revenue. All predictions are directional guidance — not the sole basis for business decisions. Platform APIs change without notice.
# Step 1: Remove skill from OpenClaw
openclaw skills remove viralevo
# Step 2: Delete local data (optional)
rm -rf ~/.openclaw/workspace/viralevo/
# Step 3: Verify
openclaw skills list
If you reinstall later without deleting Step 2, ViralEvo will resume from your existing data.
| Symptom | Fix |
|---|---|
| "Not configured" | Run node {baseDir}/scripts/onboarding.js |
| "TAVILY_API_KEY not set" | Add key to ~/.openclaw/workspace/.env |
| No topics in report | Run node {baseDir}/scripts/collect.js first |
| System check | Run python3 {baseDir}/setup.py |
| Accuracy dropping | Run python3 {baseDir}/scripts/weekly_review.py manually |
| Quick diagnosis | Run python3 {baseDir}/scripts/status.py |
ZIP package — ready to use