Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
{"openclaw":{"requires":{"env":"[Truncated]","bins":"[Truncated]","network":"[Truncated]","permissions":"[Truncated]","runtime_installs":"[Truncated]","primaryEnv":"KRYPTOGO_API_KEY"},"security":{"default_mode":"supervised","trade_confirmation":"required_by_default","autonomous_trading":"opt_in","credential_access":"environment_variables_only","credential_file_read":"setup_script_only","credential_file_read_note":"Only scripts/setup.py reads and writes ~/.openclaw/workspace/.env for initial keypair generation and address repair. All other scripts access credentials exclusively via pre-loaded environment variables.","local_signing_only":true}}}
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.
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
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>
Tier
Daily API Calls
Trading Fee
Signal Dashboard
KOL Finder
Free
100 calls/day
1%
No
No
Pro
1,000 calls/day
0.5%
Yes
Yes
Alpha
5,000 calls/day
0%
Yes
Yes
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:
Write trade details to memory/trading-journal.json with status: "OPEN"
"conservative" (skip medium risk), "moderate" (ask on medium), "aggressive" (auto-trade medium)
require_trade_confirmation
true
Set to false for autonomous mode
chains
["solana"]
Chains to scan
Safety Guardrails
Trading Limits (Hard Caps)
Limit
Default
Overridable?
Max single trade
0.1 SOL
Yes, via max_position_size
Max concurrent positions
5
Yes, via max_open_positions
Max daily trade count
20
Yes, via max_daily_trades
Price impact abort
>10%
No — always abort
Price impact warn
>5%
No — always warn
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-executionsource ~/.openclaw/workspace/.env && bash scripts/cron-examples.sh setup-default
# Autonomous mode (opt-in): auto-buys and auto-sellssource ~/.openclaw/workspace/.env && bash scripts/cron-examples.sh setup-autonomous
# Remove all cron jobs
bash scripts/cron-examples.sh teardown
Job
Interval
Default Behavior
stop-loss-tp
5 min
Report triggered conditions, do NOT auto-sell
discovery-scan
1 hour
Analyze and send recommendations, do NOT auto-buy
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)
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
Concept
Key Insight
Cluster
Group of wallets controlled by same entity
Cluster Ratio
% of supply held by clusters. ≥30% = controlled, ≥50% = high risk
Developer
Deployed the token. Highest dump risk
Sniper
Bought within 1s of creation. Sell pressure if not cleared
Smart Money
Realized profit >$100K. Accumulation often precedes price moves
Accumulation
Cluster % rising + price consolidating = bullish
Distribution
Price rising + cluster % falling = bearish
Full concepts guide: see references/concepts.md
Best Practices
Always check /agent/account first to confirm tier and quota
Always check /agent/portfolio on startup to detect existing positions
Never expose private keys in logs, messages, or CLI arguments
Validate price impact before submitting — abort >10%, warn >5%
Sign and submit promptly — blockhash expires after ~60 seconds
Persist state to memory/trading-state.json after every action
Log every trade to journal — no exceptions
Read memory/trading-lessons.md before scanning — avoid repeating known bad patterns
