Research

ChainGPT Research Skill

When a user wants to research a token, wallet, or contract — your job is to orchestrate the right sequence of MCP tools from this plugin into a coherent answer. Do not call the SDKs raw; the MCP tools are pre-tuned for credit efficiency and AI enrichment.

Core principle: one-shot if you can

If the user asks "research X" without further qualifiers, prefer the single AI-enriched call:

chaingpt_intel_token   query="X"

It composes market + risk + ChainGPT news + AI signals in one response and costs ~1 ChainGPT credit (the news fetch). This is what 80% of "research this token" prompts should do.

For a wallet:

chaingpt_intel_wallet  address="0x…"

Composes portfolio (Moralis) + per-holding risk-rating (GoPlus). Requires MORALIS_API_KEY for the portfolio scan — without it, the tool returns a setup hint pointing to https://moralis.io (free 25k req/month tier). The GoPlus risk-rating layer is always free.

When to fan out to specific tools

Use the granular tools when the user wants depth on one axis the intel call summarized.

| User intent | Tool | Notes | |---|---|---| | Live price + volume + liquidity | chaingpt_research_token | DexScreener; no key | | Which pools / pairs exist | chaingpt_research_pairs | sort by liquidity | | Trending right now | chaingpt_research_trending | boosted tokens on DexScreener | | Wallet portfolio (multi-chain) | chaingpt_wallet_balances | needs MORALIS_API_KEY for ERC-20 scan | | Wallet DeFi positions | chaingpt_wallet_positions | Aave / Uniswap / Lido / etc. | | Wallet realized P&L | chaingpt_wallet_pnl | tax-season report style | | Token safety flags | chaingpt_risk_token | GoPlus — honeypot/mintable/proxy/tax | | Buy+sell simulation | chaingpt_risk_honeypot | Honeypot.is — real simulation | | Is this address a scammer | chaingpt_risk_address | sanctions / phishing / mixer | | Source code & ABI of a contract | chaingpt_risk_contract_source | Etherscan v2 verified source | | Decode a specific transaction | chaingpt_onchain_tx | status / gas / method | | Address recent activity | chaingpt_onchain_address | last 25 txs | | Current gas prices | chaingpt_onchain_gas | safe / standard / fast | | ChainGPT-curated news | chaingpt_news_fetch | filter by token/category |

Recommended chains and graceful fallback

The plugin supports 11 chains out of the box: ethereum, base, arbitrum, optimism, polygon, bsc, avalanche, blast, linea, scroll, solana.

Some tools degrade gracefully when optional API keys are absent:

• Without MORALIS_API_KEY: chaingpt_wallet_balances returns native-coin balances only via public RPC; positions and PnL return a setup hint. Recommend the user get a free 25k-req/month key at https://moralis.io.
• Without ETHERSCAN_API_KEY: on-chain + contract-source tools work but hit a low rate limit. Recommend a free key at https://etherscan.io/myapikey.
• All risk tools work key-free.

The "research → audit" funnel

When the research surfaces a contract that looks interesting or suspicious, the next step is the existing ChainGPT audit tool:

chaingpt_risk_contract_source  address="…" chain="…"      # get source
chaingpt_audit_contract        sourceCode="…"             # AI audit

Always recommend this pair when:

• The user is about to interact with an unknown contract.
• GoPlus flagged a risk that needs a human-readable explanation.
• The contract is a proxy and the user needs to know what the implementation does.

What this skill does NOT do

• It does not execute transactions, sign anything, or move funds. All tools are read-only.
• It does not generate strategies or trading signals beyond surfacing ChainGPT's own AI signal output.
• For execution / contract deployment, see the (upcoming) chaingpt-deploy and chaingpt-trade skills.

Output style

Always lead with the most important fact: "It's a honeypot," "It's verified and looks safe," "Price is $X with $Y liquidity," etc. Then surface the supporting evidence with tool names so the user can dig deeper. End with one concrete recommended next action.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: ChainGPT-org
• Source: ChainGPT-org/chaingpt-claude-skill
• License: MIT
• Homepage: https://docs.chaingpt.org/dev-docs-b2b-saas-api-and-sdk

Install and usage instructions live in the source repository linked above.