Skip to main content
Skills are instructions to teach AI models how to do a specific task, step by step. As a dev/product, you would outline a plan to build a portfolio tracker roughly around these steps
  1. Fetch native token balances (like ETH for Ethereum)
  2. Fetch raw ERC20 token balances and USD price
  3. Format the token values by multipling the raw values by decimal places
  4. Sort tokens by the largest USD value
  5. Sum all USD values to get a net worth
Having built many of these features ourselves for Etherscan, we’ve created specific steps and common gotchas in our Skills page.

​
Skills vs MCP

Here’s an outline of these buzzwords.

MCP

A standard to fetch data, often gated with sign in/API keys, eg token balances

Skills

What to do with the data, eg step by step instructions to build a token portofolio page, formatting ETH units.
Often they would be combined, when working with a prompt like 
Analyse this address holding $HYPE
Using Skills and MCP in tandem, a workflow would look like
  1. Check Skills for instructions to analyse an address
  2. Skills outlines steps to get a user’s txns, isolate addresses interacted with, and fetch labels for these addresses
  3. Model connects to an MCP to retrieve the user txns and labels
  4. Model observes that tokens are sent to addresses labelled “Bybit 17”, and reasons this as selling
  5. Model presents a summary to the user