Imported from PinnacleCryptNG/siftra (.agents/skills/binance-leaderboard/SKILL.md). Install upstream with npx skills add PinnacleCryptNG/siftra --skill binance-leaderboard. Copyright stays with the author.

Binance Leaderboard Skill

On-chain wallet leaderboard ranking and address analysis. Query top traders by PnL, win rate, and more. Evaluate wallet quality with a 6-dimension scoring model.

Prerequisites

This skill requires the baw CLI (@binance/agentic-wallet npm package). If baw is not found:

npm install -g @binance/agentic-wallet

Verify: baw --version should print 1.6.2 or higher. If installation fails or the user doesn't have Node.js, inform them that Node.js >= 18 is required.

When to Use

Supported Chains

Command Tree

baw leaderboard
  query                     # Leaderboard query (Public, no auth)
  analyze                   # Single address analysis (6-dim score + AI overlay)
  alpha-radar              # Gem Hunter query (Private, agentSessionId)
  preset
    save                    # Save preset filters (Private)
    list                    # List preset filters (Private)
  alpha-radar-config
    save                    # Save Gem Hunter config (Private)
    list                    # List Gem Hunter config (Private)

All commands support --json for structured output.

Leaderboard Query

# Basic query — top 20 by PnL on BSC, 7d period
baw leaderboard query -c 56 -p 7d -t ALL --json

# Sort by win rate, KOL tag only
baw leaderboard query -c 56 -p 30d -t KOL --sort-by 20 --json

# Pagination (page from 0, size max 20)
baw leaderboard query -c 56 -p 7d --page 0 --size 20 --json

Public endpoint — no auth required. Returns per-address PnL, win rate, volume, trade count, token distribution, daily PNL, and top earning tokens.

Key query parameters: -c/--chain-id (required), -p/--period (7d/30d/90d), -t/--tag (ALL/KOL/MPC), --sort-by (0=PnL · 20=Win Rate · 30=Total Volume · 50=Trade Count · 60=Recent Activity · 70=Profit Rate · 80=Token Count), --order-by (0/2=Descending · 1=Ascending), --page (from 0), --size (max 20).

Full parameter and return field reference: references/cli.md

Single Address Analyze

# Analyze a wallet address — 6-dim scoring + AI archetype
# Default scans top 1000 entries
baw leaderboard analyze -c 56 -a 0xabc... --json

# Scan more entries (up to 5000) for long-tail addresses
baw leaderboard analyze -c 56 -a 0xabc... --top-n 5000 --json

Evaluates the address across 6 dimensions (winrate 25 + stability 20 + drawdown 20 + tags 15 + pnl 10 + follow_friendly 10 = 100), then applies an AI overlay (archetype + behavior_flags + ai_adjustment ±10).

Flow: Query leaderboard top 1000 (configurable via --top-n) → reverse-lookup the target address → compute scores → apply AI overlay → output rating.

Rating: ⭐⭐⭐ ≥ 80 · ⭐⭐ ≥ 65 · ⭐ ≥ 50 · ❌ < 50

If the address is not in the top N (default 1000), returns "N beyond top". Use --top-n to increase scan range up to 5000.

Full scoring model details: references/scoring.md

Gem Hunter

# Find wallets holding specific tokens
baw leaderboard alpha-radar -c 56 \
  -t 0xtoken1,0xtoken2 \
  -m 1 --json

Private endpoint — requires agentSessionId. Extra required params: -t/--tokens (comma-separated token addresses), -m/--match-count (≥ 1). Supports -p/--period, --page, --size like query.

Returns records with the same fields as leaderboard query, but topEarningTokens replaced by marchedTokens (matched tokens).

Note: The field is spelled marchedTokens (not "matched") in the CLI output.

Preset & Gem Hunter Config

# Save preset filter conditions (pass null/empty to clear all)
baw leaderboard preset save --config '[{"name":"MyPreset","period":"7d","winRateMin":50}]' --json

# List saved presets
baw leaderboard preset list --json

# Save Gem Hunter config (pass null/empty to clear)
baw leaderboard alpha-radar-config save -c 56 \
  --config '[{"uuid":"u1","name":"MyRadar","matchTokenCount":2,"tokenAddressList":[{"tokenAddress":"0xabc"}]}]' --json

# List saved configs
baw leaderboard alpha-radar-config list -c 56 --json

Core Rules

Scoring Model Overview

6-dimension model (total 100) + AI overlay (±10). Dimensions: winrate (25), stability (20), drawdown (20), tags (15), pnl (10), follow_friendly (10). Full tiered scoring tables and AI overlay rules: references/scoring.md.

Top-N Reverse Lookup

analyze queries the leaderboard's top 1000 entries (configurable via --top-n, max 5000) and reverse-looks-up the target address. If not found, returns "N beyond top" — inform the user they can increase --top-n or use binance-wallet-tracker's address list for long-tail wallets.

Preset & Config Clearing

Passing null or empty array to preset save or alpha-radar-config save clears all saved items.

Cross-Skill Scenarios

Some scenarios require both leaderboard and wallet-tracker skills:

• evaluate follow list: Use binance-wallet-tracker's tracker follow to get the user's followed addresses, then leaderboard analyze each one.
• leaderboard diff: Query leaderboard top N, then diff against binance-wallet-tracker's address list to find untracked wallets.
• batch import: Query leaderboard with filters → output address list → feed to binance-wallet-tracker's address batch command.

Write-Back Confirmation

After preset/config save operations, re-fetch via list to confirm the operation succeeded — don't assume success from the API response alone. If readback shows the saved data is missing, tell the user "Save may not have taken effect, please try again later" — never mention backend bugs or silent failures.

User-Facing Presentation

This skill serves end users, not developers. Internal field names, error codes, and CLI internals must never appear in user-facing output.

1. Term Mapping (internal → user-facing)

Raw enum values (sniper, swing, etc.) may appear in CLI syntax examples and internal lookup tables, but never in user-facing replies.

2. Never expose internal identifiers

• groupId, displayOrder — internal IDs, never shown to users
• Error codes (70001001, etc.) — translate to natural-language messages only
• Backend behavior details (e.g. preset save returning data: true without persisting) — never mention "backend bug" or "silent failure"

3. Display template hygiene

Use user-facing terms in all output. The archetype field should be translated to its Chinese term (Sniper, Swing Trader, etc.) — never show the raw English enum value. Score should be shown as a number, not as a formula.

Error Codes

Error codes are for internal lookup only — never show numeric codes or internal names to users. Translate to natural-language messages.

Display Templates

Leaderboard entry:

{addressLabel} | PnL: {realizedPnl} ({realizedPnlPercent}%) | Win Rate: {winRate}% | Trades: {totalTxCnt} | Tokens: {totalTradedTokens}

Address analysis rating:

📊 {addressLabel} Address Analysis

Rating: ⭐⭐⭐ (85/100)
Trading Style: Sniper
Behavior Patterns: High Frequency Small Amount, Nocturnal Active

Dimension Scores:
  Win Rate: 22/25 | Stability: 18/20 | Drawdown: 16/20
  Tags: 15/15 | PnL: 8/10 | Trackability: 6/10

AI Adjustment: +5 (consistent trading style, stable pattern)

Full CLI Reference

• references/cli.md — All commands with parameter tables, return field tables, and examples
• references/scoring.md — 6-dimension scoring model, AI overlay, rating standards