SKILL.md
Token Discovery
Answers: "What tokens are trending and worth a deeper look?"
CHAIN=solana
# Screen top tokens by volume
nansen research token screener --chain $CHAIN --timeframe 24h --limit 20
# → token_symbol, price_usd, price_change, volume, buy_volume, market_cap_usd, fdv, liquidity, token_age_days
# Smart money only
nansen research token screener --chain $CHAIN --timeframe 24h --smart-money --limit 20
# Search within screener results (client-side filter)
nansen research token screener --chain $CHAIN --search "bonk"
# Smart money holdings — what SM wallets are holding
nansen research smart-money holdings --chain $CHAIN --labels "Smart Trader" --limit 20
# → token_symbol, value_usd, holders_count, balance_24h_percent_change, share_of_holdings_percent
# Nansen indicators for a specific token
TOKEN=<address>
nansen research token indicators --token $TOKEN --chain $CHAIN
# → risk_indicators, reward_indicators (each with score, signal, signal_percentile)
# Flow intelligence — only use for promising tokens from screener/indicators above
nansen research token flow-intelligence --token $TOKEN --chain $CHAIN
# → net_flow_usd per label: smart_trader, whale, exchange, fresh_wallets, public_figure
# Nansen Score Top Tokens — "what should I buy?" (public endpoint, any authenticated API key)
# Use this FIRST for discovery, then drill into individual tokens with `indicators` above
nansen research token top-tokens --limit 25
nansen research token top-tokens --market-cap largecap --limit 10
# → chain, token_address, token_symbol, performance_score, risk_score,
# per-indicator contributions, market_cap_group, latest_date, last_trigger_onScreener timeframes: 5m, 10m, 1h, 6h, 24h, 7d, 30d
Indicators: score is "bullish"/"bearish"/"neutral". signal_percentile > 70 = historically significant. Some tokens return empty indicators — not an error.
Top tokens — Nansen Score field reference
Results are pre-filtered to performance_score >= 15 server-side and returned sorted by:
performance_scoreDESCmarketcapgrouppriority (largecap → midcap → lowcap)risk_scoreDESC- 24h volume DESC
So row 0 is always the strongest candidate for the filter you applied — no client-side ranking needed.
Market cap buckets (used in both the sort priority and the --market-cap filter):
lowcap: market cap < $100Mmidcap: market cap $100M – $1Blargecap: market cap > $1B
Every contribution is ternary — exactly one of {negative, 0, positive} per field. No partial values. Zero means "indicator didn't apply to this token" (out of scope), not "indicator was neutral".
Performance Score (Alpha — "likely to outperform BTC over 7–30d") Range: -60 to +75 (arithmetic bounds; live max is closer to +45 since no single token hits every positive indicator simultaneously). Buy threshold: >= 15. Sum of the five *_performance fields below.
| Field | Contribution | Trigger | What the underlying indicator measures |
|---|---|---|---|
pricemomentumperformance |
+30 / 0 | upstream score bullish → +30 |
Price momentum, scored against separate thresholds for large-cap vs. low/mid-cap tokens. |
chainfeesperformance |
+30 / 0 | bullish (30-day fee growth > +1%) → +30 |
30-day spending momentum on network fees (geometric mean of daily returns). Only tracked for a handful of L1 native tokens (e.g. ETH, TRX, AVAX, RON); always 0 for every other token. |
tradingrangeperformance |
+15 / 0 | bullish (price breaks above resistance in an uptrend) → +15 |
14-day price trend combined with position vs. nearest support/resistance. In practice fires mostly on established tokens that have well-defined levels — can fire at any market cap, but is rare for new / low-liquidity tokens. |
chaintvlperformance |
0 / -35 | bearish (composite TVL growth < 0) → -35 |
TVL momentum composite signal. Only non-zero for chains / L2s whose TVL is tracked. No positive path exists — the field only deducts. |
protocolfeesperformance |
0 / -25 | bearish (14-day fee growth < -3%) → -25 |
14-day protocol fee momentum. Only non-zero for tokens backed by protocols with measurable fee revenue. No positive path — deduction only. |
Risk Score (Safety — "filters falling knives / dangerous setups") Range: -60 to +80 (arithmetic bounds). Safety threshold: > 0 (positive = safer, negative = riskier). Sum of the four *_risk fields below. For every risk field: upstream score low → positive contribution, high → negative contribution, medium/missing → 0.
| Field | Contribution | What the underlying indicator measures |
|---|---|---|
btcreflexivityrisk |
+40 / -20 | Rolling 5-event median ratio of token drop to BTC drop on days BTC falls >3%. Ratio ≤ 1 → low → +40 (token holds up as well as or better than BTC on drawdowns). Ratio > 1 → high → -20 (token drops harder than BTC). Skipped for stablecoins and tokens with <$1M 24h volume. |
liquidity_risk |
+20 / -20 | Ratio of on-chain liquidity to market cap (totalliquidityusd / marketcapusd). Higher ratio → low → +20 (deep books relative to cap). Very thin ratio → high → -20. |
concentration_risk |
+10 / -10 | Top-10 holder concentration as a fraction of supply. < 0.12 → low → +10 (well-distributed). > 0.55 → high → -10 (whale-concentrated). |
inflation_risk |
+10 / -10 | EMA of daily token supply inflation rate. Negative / near-zero → low → +10 (stable or deflationary supply). Strongly positive → high → -10 (high dilution). Only evaluated for tokens >= $100M market cap. |
Other response fields:
marketcapgroup:lowcap/midcap/largecap— see thresholds above.latest_date: ISO datetime of the most recent indicator refresh for this token.lasttriggeron: ISO datetime of the most recent trigger across contributing indicators (MAX aggregate — individual indicators may be days-to-months stale even when this looks fresh). Useindicatorson a specific token to audit per-indicator ages.
Stablecoins rank high but aren't picks. USDC, USDT, DAI, FDUSD and similar score well on chainfees + liquidity indicators but aren't what "what should I buy" means. Filter them out of the shortlist using the canonical whitelist at nansen-dbt-ch-tokens/seeds/stablecoinsfor_indicator.csv before drilling into indicators.
Typical workflow: start with top-tokens for a shortlist → drop stablecoins → run indicators on the top 3–5 to inspect individual signals and their signal_percentile → flow-intelligence only on the finalists to confirm SM conviction.
Field meanings and contribution mappings above are sourced from nansen-dbt-ch-tokens/models/indicators/apinansenscoreindicatorsalltokenslatest.sql and per-indicator model yml files. Sign conventions and live value ranges were validated against production ClickHouse data.
Flow intelligence is credit-heavy. Use it to confirm SM conviction on tokens that already look promising from screener + indicators, not as a first pass on every token.