KryptoGO Meme Trader Agent Skill

Overview

This skill enables an AI agent to analyze and trade meme coins through the KryptoGO platform, combining deep on-chain cluster analysis with trade execution.

Analysis (multi-chain: Solana, BSC, Base, Monad): wallet clustering, accumulation/distribution detection, address behavior labels, network-wide accumulation signals (Pro/Alpha tier).

Trading (Solana only): portfolio monitoring with PnL tracking, swap execution via DEX aggregator, local transaction signing (private key never leaves the machine).

Default mode is supervised — all trades require user confirmation. Autonomous trading is available as opt-in. See references/autonomous-trading.md for autonomous mode, cron setup, and learning system details.

When to Use

• User asks to analyze a meme coin or token on Solana/BSC/Base/Monad
• User asks to trade, buy, or sell tokens
• User asks to scan for trending tokens or market opportunities
• User asks to monitor portfolio positions or check PnL
• Cron-triggered periodic portfolio monitoring and signal scanning

When NOT to Use

• BTC/ETH/major L1 macro analysis, NFTs, cross-chain bridging, non-DEX transactions, non-Solana trading

Setup Flow

1. Get API Key

1. Go to kryptogo.xyz/account and create an API key
2. Add to ~/.openclaw/workspace/.env:

   echo 'KRYPTOGO_API_KEY=sk_live_YOUR_KEY' >> ~/.openclaw/workspace/.env && chmod 600 ~/.openclaw/workspace/.env
   

Do NOT paste your API key directly in chat. Always set secrets via .env file.

2. Generate Agent Wallet

python3 scripts/setup.py

Creates a Solana keypair, saves to .env with chmod 600, prints public address to fund.

3. Fund the Wallet

Send SOL to the agent's public address (minimum 0.1 SOL).

Security Rules

• NEVER print, log, or include private keys in any message or CLI argument
• NEVER accept secrets pasted directly in chat — instruct users to set them in .env
• NEVER use the Read tool on .env — load credentials via source command only
• Runtime scripts do NOT read .env directly — all credentials are accessed via environment variables only, which must be pre-loaded by the caller (source ~/.openclaw/workspace/.env)
• Exception: scripts/setup.py reads and writes .env for initial keypair generation and address repair — this is the only script that touches credential files
• Private key stays in memory only during local signing — never sent to any server

Authentication

All endpoints require: Authorization: Bearer sk_live_<48 hex chars>

Agent Behavior

Session Initialization

On every session start (including heartbeat/cron), the agent MUST load credentials BEFORE running any scripts:

source ~/.openclaw/workspace/.env

This is REQUIRED — scripts do not read .env directly. All credentials are accessed via environment variables only.

Default Mode: Supervised

By default, the agent operates in supervised mode: it analyzes tokens, presents recommendations, and waits for user approval before executing any trade. Stop-loss/take-profit conditions are reported to the user but not auto-executed.

To enable autonomous trading, set require_trade_confirmation: false in preferences. See references/autonomous-trading.md for full details.

Persistence (CRITICAL)

IMMEDIATELY after submitting a transaction, the agent MUST:

1. Write trade details to memory/trading-journal.json with status: "OPEN"
2. Include: token_symbol, token_address, entry_price, position_size_sol, tx_hash, timestamp

User Preferences

Store in memory/trading-preferences.json:

Safety Guardrails

Trading Limits (Hard Caps)

If any limit is hit, the agent must stop and notify the user.

Credential Isolation

Runtime scripts in this skill do NOT read .env files directly. All credentials are accessed via environment variables only, which must be pre-loaded by the caller (source ~/.openclaw/workspace/.env). This ensures no runtime script can independently access or exfiltrate credential files.

Exception: scripts/setup.py reads and writes .env — it loads existing keys to avoid regeneration, backs up .env before changes, and writes new keypair entries. This is the only script that touches credential files, and it runs only during initial setup or explicit --force regeneration.

Automated Monitoring (Cron)

Quick Setup

# Supervised mode (default): analysis + notifications, no auto-execution
source ~/.openclaw/workspace/.env && bash scripts/cron-examples.sh setup-default

# Autonomous mode (opt-in): auto-buys and auto-sells
source ~/.openclaw/workspace/.env && bash scripts/cron-examples.sh setup-autonomous

# Remove all cron jobs
bash scripts/cron-examples.sh teardown

For full cron configuration, manual setup, heartbeat alternative, and monitoring workflow details, see references/autonomous-trading.md.

On-Chain Analysis Framework (7-Step Pipeline)

Step 1: Token Overview & Market Cap Filter

/token-overview?address=<mint>&chain_id=<id> — get name, price, market cap, holders, risk_level. Skip if market cap < min_market_cap.

Step 2: Cluster Analysis

/analyze/<mint>?chain_id=<id> — wallet clusters, top holders, metadata.

• ≥30-35% = "controlled" — major entity present
• ≥50% = high concentration risk
• Single cluster >50% → skip (rug pull risk)

Free tier limitation: Cluster analysis only returns the top 2 clusters. To see full cluster data, upgrade at kryptogo.xyz/pricing.

Step 3: Cluster Trend (Multi-Timeframe)

/analyze-cluster-change/<mint> — cluster_ratio + changes across 15m/1h/4h/1d/7d.

Core insight: Price and cluster holdings DIVERGING is the key signal.

• Rising price + falling cluster % = distribution (bearish)
• Falling price + rising cluster % = accumulation (bullish)

Step 4: Address Labels + Sell Pressure Verification

1. /token-wallet-labels → identify dev/sniper/bundle wallets
2. /balance-history for each risky address → check if still holding
3. Compute risky_ratio = active risky holdings / total cluster holdings

4. 30% = high risk, 10-30% = medium, <10% = low

Labels represent behavioral history, not current holdings. Always verify via /balance-history.

Step 5: Deep Dive (Optional)

/balance-history, /balance-increase/<mint>, /top-holders-snapshot/<mint>, /analyze-dca-limit-orders/<mint>, /cluster-wallet-connections

Step 6: Decision

Apply Bullish Checklist from references/decision-framework.md.

Step 7: Execute Trade

Use scripts/swap.py for execution — handles wallet_address injection, error checking, and journal logging.

source ~/.openclaw/workspace/.env && python3 scripts/swap.py <token_mint> 0.1
source ~/.openclaw/workspace/.env && python3 scripts/swap.py <token_mint> <amount> --sell

API Quick Reference

Full request/response details: see references/api-reference.md

Multi-Chain Support

Error Handling

Operational Scripts

All scripts require credentials to be pre-loaded: source ~/.openclaw/workspace/.env before running.

source ~/.openclaw/workspace/.env && bash scripts/portfolio.sh              # Portfolio check
source ~/.openclaw/workspace/.env && bash scripts/trending.sh               # Trending tokens
source ~/.openclaw/workspace/.env && bash scripts/analysis.sh               # Full analysis dashboard
source ~/.openclaw/workspace/.env && python3 scripts/swap.py <mint> 0.1     # Buy
source ~/.openclaw/workspace/.env && python3 scripts/swap.py <mint> <amt> --sell  # Sell
source ~/.openclaw/workspace/.env && bash scripts/test-api.sh               # API connectivity test

Learning & Adaptation

The agent improves over time by recording trades, analyzing outcomes, and adjusting strategy. Every trade is logged to memory/trading-journal.json, losses trigger post-mortems, and periodic reviews propose parameter changes.

For full details on the learning system, trade journal format, post-mortem process, and strategy reviews, see references/autonomous-trading.md.

Core Concepts

Full concepts guide: see references/concepts.md

Best Practices

1. Always check /agent/account first to confirm tier and quota
2. Always check /agent/portfolio on startup to detect existing positions
3. Never expose private keys in logs, messages, or CLI arguments
4. Validate price impact before submitting — abort >10%, warn >5%
5. Sign and submit promptly — blockhash expires after ~60 seconds
6. Persist state to memory/trading-state.json after every action
7. Log every trade to journal — no exceptions
8. Read memory/trading-lessons.md before scanning — avoid repeating known bad patterns

File Structure

kryptogo-meme-trader/
├── SKILL.md                       ← You are here
├── package.json
├── .env.example
├── references/
│   ├── api-reference.md           ← Full API docs
│   ├── concepts.md                ← Core concepts
│   ├── decision-framework.md      ← Entry/exit strategies
│   └── autonomous-trading.md      ← Autonomous mode, cron, learning system
├── scripts/
│   ├── setup.py                   ← First-time setup
│   ├── cron-examples.sh           ← Cron configurations
│   ├── portfolio.sh / trending.sh / analysis.sh / test-api.sh
│   ├── swap.py                    ← Swap executor
│   └── trading-preferences.example.json
└── examples/
    ├── trading-workflow.py
    └── deep-analysis-workflow.py