Smart-Social desk

Agent token 7. The Smart-Social desk reads where crypto attention is rotating — using Elfa's smart-account data (curated influential accounts, not retail noise). It quantifies attention magnitude, not direction.

What it reads

firm/organization.py · tool_smartsocial() → firm/elfa_client.py:

• Trending narratives — Elfa trending-tokens (24h mindshare): which tokens/narratives are heating up.
• Per-asset top-mentions — engagement-weighted mentions for the traded ticker.
• Per-asset news — recent token-news items.
• Attention magnitude — total views / mention count, reported as measured each cycle, so the gap between BTC and a smaller NEAR-ecosystem asset is visible, not smoothed over.

How attention becomes context

Output (shape)

{
  "asset": "BTC",
  "asset_in_top_trending": true,
  "asset_attention_24h": { "smart_mentions": 5, "total_views": 1000000, "news_items": 4 },
  "asset_top_mentions": ["…"],
  "asset_news": ["…"],
  "top_narratives_24h": ["…"]
}

Honest caveats

• Attention is context, not direction. A loud asset can dump; a quiet one can rip. The desk supplies how much an asset is in the conversation, and the PM decides what that's worth.
• Elfa's AI trade-analysis chat (/v2/chat) is paid-tier — our key returns 403 scope: chat. So we use the data endpoints (which our key reaches) and encode Elfa's trading framework (no mid-range chop, dynamic-stall exit, event-risk) directly in the firm's gates.
• Mindshare for smaller NEAR-ecosystem assets can be thin — the desk reports that honestly rather than inventing a narrative.