GoldRush Skills: Structured Knowledge for AI Agents

Covalent just shipped something called GoldRush Agent Skills, which are four structured documentation packages designed to give AI agents the context they need to call GoldRush APIs correctly. The idea is that instead of hoping your agent hallucinates the right endpoint, you feed it a structured knowledge package that describes exactly how the API works, what parameters it expects, and what the response looks like.

If you have been doing any vibecoding with blockchain data, you have probably run into the problem where the LLM confidently invents an endpoint that does not exist, or gets the argument order wrong, or just returns something plausible-looking that is completely wrong. Skills are meant to address that by giving the model grounded, structured context up front.

The GoldRush CLI already exposes a lot of this directly in the terminal. Some examples of what is available:

“`bash

goldrush balances <chain> <address>

goldrush transfers <address> <chain>

goldrush watch <address> <chain>

goldrush gas [chain]

goldrush traders <token\_address> <chain>

“`

The Skills packages extend this into agent workflows, so the same underlying API knowledge is available whether you are running CLI commands manually or wiring up an LLM agent to do it programmatically.

A few honest observations after looking at this:

– Setup does require an API key and auth step (`goldrush auth`) before anything works, so there is some friction before you see results.

– The Skills packages are structured for agent consumption, which means they are more useful if you are already building an agent pipeline. If you just want quick terminal lookups, the CLI on its own is probably sufficient.

– This is a product from Covalent, not a solo side project, so the roadmap and supported chains depend on what they prioritize. The `goldrush chains` command will show you what is currently live.

– Agent tooling here is model-agnostic. The structured knowledge packages are not tied to a specific LLM.

Blog post with more detail on the Skills packages: https://goldrush.dev/blog/goldrush-skills-structured-knowledge-for-ai-agents/