diff --git a/skills/nansen-convergence-scanner/REFERENCE.md b/skills/nansen-convergence-scanner/REFERENCE.md
new file mode 100644
index 00000000..1c55da9e
--- /dev/null
+++ b/skills/nansen-convergence-scanner/REFERENCE.md
@@ -0,0 +1,61 @@
+# Convergence Scanner — Reference
+
+## Why convergence matters
+
+Individual wallet tracking shows one actor's opinion. Convergence detects when **multiple unrelated actors independently reach the same conclusion** — a qualitatively stronger signal than any single whale trade.
+
+A Fund, a Smart Trader, and an independent whale all accumulating the same token within the same window is not coordination — it's market wisdom emerging from independent analysis.
+
+## Independence Protocol
+
+Run Phase 3 for every pair of addresses on the same token. Skip the token if ANY pair fails independence:
+
+1. `profiler related-wallets` on address A — if B appears, **same entity**
+2. `profiler compare` on A,B — if `shared_counterparties >= 3`, **likely same entity**
+3. `profiler labels` on both — if identical labels AND same first funder, **same entity**
+
+**Independent = passes all three tests.**
+
+## Label diversity scoring
+
+Convergence is stronger when entities have different label types:
+
+| Mix | Multiplier | Rationale |
+|-----|-----------|-----------|
+| Fund + Smart Trader | 1.5x | Different strategies, same conclusion |
+| Fund + Whale | 1.3x | Informed + size |
+| Smart Trader + Fresh Wallet | 1.0x | Fresh could be noise |
+| Same label on both | 0.8x | Possible coordination despite independence tests |
+
+## Multi-chain convergence
+
+When the same token trades on multiple chains (e.g., ETH on ethereum + arbitrum + base), extend the scan:
+
+```bash
+for CHAIN in ethereum base arbitrum; do
+  nansen research smart-money netflow --chain $CHAIN --limit 20 --fields token_symbol,token_address,net_flow_usd
+done
+```
+
+Cross-chain convergence (SM buying on ethereum AND base) is an even stronger signal than same-chain convergence.
+
+## False positives to filter
+
+- **Stablecoins** — USDC/USDT/DAI will always show multiple SM addresses. Exclude them.
+- **Wrapped natives** — WETH/WSOL are operational, not directional. Exclude them.
+- **Airdrop farming** — Multiple addresses buying a new token on launch day. Check if addresses were created recently (`profiler labels` → "Fresh Wallet"). If all are fresh → likely farm, not convergence.
+- **Market maker flows** — Addresses labeled "Market Maker" are providing liquidity, not taking directional bets. Exclude them from convergence counting.
+
+## Conviction boost
+
+When convergence is confirmed, the implied conviction of related signals should increase:
+- **Moderate** (2 entities): +10 to any prior conviction score on this token
+- **Strong** (3+ entities): +15, capped at 95
+- **Very strong** (3+ with label diversity): +20, capped at 95
+
+## Cost warnings
+
+- Each candidate pair costs 3-4 API calls for independence verification
+- A token with 5 SM addresses = 10 pairs = up to 40 calls — cap at 3 pairs per token
+- Use `--fields` on Phase 1 calls to reduce response size
+- `profiler compare` may return sparse data for wallets with few on-chain interactions
diff --git a/skills/nansen-convergence-scanner/SKILL.md b/skills/nansen-convergence-scanner/SKILL.md
new file mode 100644
index 00000000..633a3528
--- /dev/null
+++ b/skills/nansen-convergence-scanner/SKILL.md
@@ -0,0 +1,104 @@
+---
+name: nansen-convergence-scanner
+description: "Detect when multiple independent smart money entities accumulate the same token. The highest-alpha signal: unrelated actors independently reaching the same conclusion."
+metadata:
+  openclaw:
+    requires:
+      env:
+        - NANSEN_API_KEY
+      bins:
+        - nansen
+    primaryEnv: NANSEN_API_KEY
+    install:
+      - kind: node
+        package: nansen-cli
+        bins: [nansen]
+allowed-tools: Bash(nansen:*)
+---
+
+# Convergence Scanner
+
+**Answers:** "Are multiple independent smart money actors converging on the same token?"
+
+Chain: `0x` → `--chain ethereum` (also base, arbitrum, optimism, polygon). Base58 → `--chain solana`.
+
+## Workflow
+
+```bash
+CHAIN=solana  # or ethereum, base
+
+# ── Phase 1: Collect SM activity ────────────────────────────
+# 1. Smart money netflow — who's accumulating?
+nansen research smart-money netflow --chain $CHAIN --limit 20 --fields token_symbol,token_address,net_flow_usd
+# 2. SM DEX trades — who's buying on-chain right now?
+nansen research smart-money dex-trades --chain $CHAIN --limit 30 --fields token_symbol,token_address,maker,taker,value_usd
+
+# ── Phase 2: Group by token ────────────────────────────────
+# From results, find tokens with 2+ DISTINCT wallet addresses.
+# These are convergence candidates.
+
+# ── Phase 3: Verify independence ────────────────────────────
+# For each candidate token with addresses ADDR_A and ADDR_B:
+ADDR_A=<first_address>
+ADDR_B=<second_address>
+
+# 3a. Check if they're related (same entity = NOT convergence)
+nansen research profiler related-wallets --address $ADDR_A --chain $CHAIN
+# If ADDR_B appears in ADDR_A's related wallets → same entity, skip.
+
+# 3b. Compare directly — shared counterparties + tokens
+nansen research profiler compare --addresses "$ADDR_A,$ADDR_B" --chain $CHAIN
+# If shared_counterparties >= 3 → likely same entity, skip.
+
+# 3c. Label both — different entity types strengthens signal
+nansen research profiler labels --address $ADDR_A --chain $CHAIN
+nansen research profiler labels --address $ADDR_B --chain $CHAIN
+
+# ── Phase 4: Enrich convergence signal ──────────────────────
+TOKEN=<token_address>
+# 4a. Token fundamentals
+nansen research token info --token $TOKEN --chain $CHAIN
+# 4b. Flow breakdown by label (Fund, Smart Trader, Whale, etc.)
+nansen research token flow-intelligence --token $TOKEN --chain $CHAIN
+```
+
+## Independence test
+
+Two addresses are **independent** when ALL of these hold:
+- Address B is NOT in address A's `related-wallets`
+- `profiler compare` shows < 3 shared counterparties
+- They have different labels (Fund vs Smart Trader strengthens signal)
+
+If any test fails → same entity, not convergence.
+
+## Signal strength
+
+| Independent entities | Strength | Action |
+|---------------------|----------|--------|
+| 2 | Moderate | Flag for review |
+| 3+ | Strong | High-priority alert |
+| 3+ with different label types | Very strong | Highest conviction signal |
+
+## Output format
+
+For each convergence signal, report:
+
+- **Token**: symbol, address, chain
+- **Entity count**: number of independent actors
+- **Actors**: address, label, direction (buy/sell), value
+- **Independence proof**: no related-wallets link, < 3 shared counterparties
+- **Flow context**: token flow-intelligence breakdown (which label categories are buying)
+
+## Credit cost
+
+Per convergence scan: ~6-12 CLI calls.
+- Phase 1: 2 calls (netflow + dex-trades)
+- Phase 3: 3-4 calls per candidate pair (related-wallets + compare + labels)
+- Phase 4: 2 calls per confirmed signal (info + flow-intelligence)
+
+Keep `--limit` low to control costs. Most scans find 0-3 convergence candidates.
+
+## Source
+
+- npm: https://www.npmjs.com/package/nansen-cli
+- GitHub: https://github.com/nansen-ai/nansen-cli