Share post

Share post

Blog: Product

Omnata CoCo skill upgrades: more capabilities with improved trust

Omnata CoCo skill upgrades: more capabilities with improved trust

New features |

Cortex Code

By: Chris Chandler

4 November 2026

  • Securely configure Omnata connections

  • Build and run syncs end to end

  • Run detailed troubleshooting

These are just some of the new capabilities of the Omnata skill for CoCo (Snowflake Cortex Code).

Chat based workflows for Omnata are now a near-parity alternative to the UI, providing a more nuanced experience that helps you make configuration decisions as well as implement them.

Configure pipelines end to end from CoCo

Tell CoCo you've just installed a plugin, and it takes it from there. It creates the connection, configures the sync, agrees where the data should land and runs the first sync, carrying on from one step to the next without being prompted.

This seamless and powerful orchestration is enabled by Omnata's native stored proc and functions interface. Everything you can do in the Omnata UI can be accessed via code. Open it in the UI afterwards and carry on, or start in the UI and ask CoCo to pick it up. Use whichever suits the task.

What else is new

Configure at scale. Got a source with hundreds of tables? Hand CoCo a list or a file and it configures every stream in one pass.

Test before production. CoCo can set up a test branch on a sandbox connection, so outbound changes are proven before they reach your production systems.

Decide where data lands. Design a storage location template, preview the exact table and view names, and check for collisions before anything is created.

Always current. The skill reads each plugin's strategies and capabilities from the apps installed in your account, so it works from what you actually have.

Sharper troubleshooting. When a sync fails, CoCo reads the run history, record errors and event table stack traces, checks whether the warehouse was over capacity at the time, and tells you whether the problem sits in Snowflake, in the source app or in Omnata. Then it explains what to change. Need us? It assembles the support ticket for you.

More guidance than a form can give

A form shows you the options. CoCo can explain them.

Ask why a stream defaults to a particular sync strategy, what a storage location template will actually create, or whether you need a test branch for the change you're about to make, and you get an answer in the context of your own account. The skill carries the detail that's hard to fit into a UI: what each setting does, sensible defaults, the trade-offs between options, and the gotchas for each plugin.

And when something goes wrong, CoCo doesn't stop at the Omnata error. It brings in what it knows about the app on the other end, whether that's a Salesforce validation rule or a HubSpot rate limit, and tells you what to change there too.

Improved trust and security

Users are right to be cautious introducing third party skills to CoCo, which inherits the rights of the user with potentially broad access to Snowflake and the local environment. We've followed the industry leaders in implementing safety practices for the Omnata Admin skill.

Every release is now scanned with NVIDIA SkillSpector. Omnata scans every release of the skill with NVIDIA SkillSpector, the open-source scanner NVIDIA uses to vet agent skills for its own verified catalog. It checks for prompt injection, data exfiltration and unsafe behaviour. The current release rates SAFE, with every file fully analysed. The few findings we accepted are published in the repo, each with its reason.

Every change declared up front. The skill opens with a Warnings section listing everything it can do that's costly, hard to reverse, or sends data outside Snowflake. Before any of those steps, CoCo lists exactly what will be affected and waits for your yes. Deleting a sync goes further: you type its name to confirm.

Open and pinnable. The skill is Apache 2.0 licensed and every release is tagged, so you can review a version, install exactly that one, and update when you choose.

Handle credentials safely

Connecting an app means providing an API key, token or key file at some point. The skill gives you safe ways to do it, and never asks you to paste a secret into the chat.

Enter it in the Omnata app (recommended). CoCo sets up the connection, then hands you one step: open the connection in the Omnata app and enter the credential there. CoCo never sees it, and it isn't written into the SQL that lands in your query history. CoCo then picks up again and finishes the setup.

Or let CoCo securely handle it, with transparency on the risks. If you'd rather stay in CoCo, you can provide the credential through SQL. Before you choose, CoCo explains that the value will be recorded in Snowflake query history, and recommends rotating it once the connection is set up.

Either way, if a secret does end up in the chat by mistake, CoCo won't repeat it and will suggest you rotate it.

Why isn't the skills just shipped with the app?

In short, Snowflake doesn't allow it.

Skills with native apps must be coupled with an agent that has limited access to the account. Unfortunately, this inhibits a lot of the value of the skill in CoCo.

Omnata runs as a Snowflake Native App, inside a deliberately restricted sandbox. That's good for security, but it means an app can't investigate itself. Many sync problems only make sense with your wider account in view: a missing grant, a suspended warehouse, a network rule. A skill running in your agent, with your role, has that view. It can only do what you're already allowed to do, and everything it does shows up as you in your query history.

Get started

  1. Read our docs

  2. Install the skill from the omnata-labs github repo

subscribe

Deliver the goods to your inbox

subscribe

Deliver the goods to your inbox