Trend Analysis

trend-analysis

> Priority: P1 (runs weekly) > Schedule: Monday 08:00 UTC > Data sources: CoinGecko, GitHub, Google Trends proxy, DeFiLlama, social APIs > Output: Trend report in memory/research/trends/

Instructions

You are executing the trend-analysis skill for the Researcher agent.

Step 1: Load Previous Trends

Read memory/research/trends/latest.json to understand:

• Previously identified trends and their scores
• Trend trajectories (rising, peaking, declining, dead)
• Last analysis timestamp

Step 2: Gather Multi-Source Signals

A. DeFi TVL Trends:

curl -s "https://api.llama.fi/v2/historicalChainTvl/Solana"
curl -s "https://api.llama.fi/protocols"

• Track 7d and 30d TVL changes by protocol and category
• Flag categories with >20% TVL growth in 7d

B. Token Category Performance:

curl -s "https://api.coingecko.com/api/v3/coins/categories"

• Identify top-performing categories by 7d market cap change
• Flag categories with >15% gain and >$100M market cap

C. GitHub Developer Activity:

curl -s -H "Authorization: token ${GITHUB_TOKEN}" "https://api.github.com/search/repositories?q=stars:>100+pushed:>{7_days_ago}&sort=stars&order=desc&per_page=30"

• Track trending repos in crypto/DeFi/AI categories
• Identify new repos with rapid star growth

D. Social Volume:

• Check memory/research/social-signals.json for trending topics
• Cross-reference with DexScreener trending:

curl -s "https://api.dexscreener.com/token-boosts/top/v1"

Step 3: Trend Scoring

For each identified trend, compute a composite score:

| Signal | Weight | Metric | |--------|--------|--------| | TVL growth | 25% | 7d % change in category TVL | | Token performance | 20% | 7d % price change of top tokens | | Developer activity | 20% | New repos + commit velocity | | Social volume | 20% | Mention count + sentiment | | Institutional signals | 15% | Funding rounds + partnerships |

Score each 0.0-1.0, then compute weighted average.

Step 4: Classify Trend Stage

For each trend, determine lifecycle stage:

• Emerging (score 0.3-0.5): Early signals, low awareness
• Growing (score 0.5-0.7): Accelerating adoption, increasing social volume
• Peaking (score 0.7-0.9): Maximum hype, potential overvaluation
• Declining (score dropping 2 consecutive weeks): Narrative fatigue

Compare to previous week's scores to determine trajectory (accelerating, stable, decelerating).

Step 5: Generate Report

{
  "report_date": "2024-01-15",
  "trends": [
    {
      "name": "Liquid Restaking",
      "score": 0.78,
      "stage": "growing",
      "trajectory": "accelerating",
      "key_signals": [
        "EigenLayer TVL +45% in 7d",
        "3 new LRT protocols launched",
        "Vitalik blog post on restaking"
      ],
      "top_tokens": ["EIGEN", "ETHFI", "REZ"],
      "top_protocols": ["EigenLayer", "EtherFi", "Renzo"],
      "risk_factors": ["Smart contract risk", "Circular dependency"],
      "relevance_to_solana": "Jito restaking gaining traction",
      "actionable": true
    }
  ],
  "new_trends": ["trends identified for the first time"],
  "dead_trends": ["trends that dropped below 0.2 score"],
  "meta": {
    "total_trends_tracked": 15,
    "data_sources_used": 5,
    "confidence": 0.8
  }
}

Step 6: Save and Distribute

1. Write report to memory/research/trends/{YYYY-MM-DD}.json
2. Update memory/research/trends/latest.json with current state
3. Post top 3 trends to analyst mesh
4. Send notification summary via ./notify

Anti-Patterns

• Do NOT chase micro-trends that last < 1 week. Focus on sustained movements.
• Do NOT confuse price pumps with genuine trends. Look for fundamental backing.
• Do NOT overweight social signals. They are noisy and manipulable.
• Do NOT present trends as investment advice. Report data, not opinions.

Exit Codes

• SKILL_OK — trend analysis complete, N trends identified
• SKILL_PARTIAL — some data sources unavailable
• SKILL_EMPTY — no significant trend changes detected
• SKILL_FAIL — critical failure

Output

Commit message format: researcher: trend-analysis — {N} trends tracked, {M} new [{top_trend}]

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: stoaaadev
• Source: stoaaadev/stoa
• License: MIT
• Homepage: https://x.com/stoaframework

Install and usage instructions live in the source repository linked above.