How the agent picks skills
The agent works out what research a question needs, runs those skills on live data, and writes the answer from their results.
Glimz has two parts that work together. Skills do the research. The agent decides which skills a question needs, then explains what they found.
Skills are deterministic
A skill is one specific piece of research, done the same way every time. Given the same token and the same chain data, it returns the same result. Each skill reads from named sources, such as GoPlus Security, Rugcheck or a DEX aggregator, and returns its evidence with the result.
The agent does not make up numbers. It reads them from skill results.
From question to answer
- Read the question. The agent works out the subject (a token, a wallet, a chain) and what you want to know.
- Find the chain. From the address and where the token trades, or from the chain you named.
- Pick the skills. A safety question needs different research than "why did it move?".
- Run them on live data. You see each skill as it runs.
- Write the answer. The agent sums up the results in plain language and attaches the evidence.
Typical picks
| You ask | Skills the agent usually runs |
|---|---|
| "Should I buy this?" | Buy or not (which runs the safety skills, Token snapshot and Market activity) |
| "Should I copy this wallet?" | Copy or not (the copy check) |
| "Which wallets should I copy?" | Wallets to copy |
| "Buy $50 of this." | Buy or not, then a prepared trade for you to sign (Trading) |
| "Is this safe?" | Contract scan, Holder map, Liquidity lock, Exit check, Deployer history |
| "Why did it pump?" | Token snapshot, Market activity |
| "Who launched this?" | Deployer history, Launch receipt (Ethereum) |
| "Did insiders get in early?" | Insider / early-buyer check, Holder map |
| "Is this the real one?" | Copycat finder, Token snapshot |
| "What's new on Base?" | New launches |
| "Tell me about this wallet." | Wallet profile |
| "What is this wallet buying?" | Wallet trades |
| "Who is making money on this?" | Top traders |
| "Where is the money going? What's next?" | Money flow (then Buy or not on the top coins) |
These are typical, not fixed. The agent picks per question, so a follow-up can add a skill the first answer didn't need.
When a skill can't run
Some skills don't run on every chain yet (see Supported chains), and a source can be slow or down. When that happens, the agent tells you which skill didn't run and why. It does not fill the gap with a guess.
The token page
When you paste a token address, Glimz runs the token skills available on that chain and builds the verdict. That is what you see on the token page: the verdict panel and the detail tabs.