Supper Skills

Most data work doesn't start as a workflow.

It starts with a question.

You ask Supper for a cut of your sales pipeline. Then another. You refine the customer population, calculate a new metric, pull in another piece of data, manipulate the results, and adjust the analysis based on what you find.

Eventually, you get exactly what you need: a list of churned customers, a customer health score, a risk model, or another analysis that reflects the way your business actually works.

The first time through, that process is exploration.

The second time, it's repetition.

Today, we're introducing Supper Skills to make that complex data work reusable.

What are Supper Skills?

Supper Skills are reusable frameworks that guide Supper's agent through complex, multi-step data workflows.

A Skill captures the process behind an analytical task so the agent can use that process again in the future.

Depending on the workflow, a Skill can include:

• Required inputs
• Instructions and analytical steps
• Success criteria
• Reasoning loops
• References to other Skills
• Relevant parts of the Supper semantic model
• Algorithms or calculations
• The tools required to complete the task

That means a Skill isn't simply a saved prompt or a recorded sequence of clicks.

It gives the agent a framework for accomplishing a goal.

Some Skills can be relatively deterministic. Others can leave room for the agent to reason, investigate, and choose the right next step based on the data it finds.

The goal is not to remove reasoning from complex analysis.

It's to make successful reasoning repeatable.

The problem: getting the answer once isn't enough

As customers spend more time with Supper, their questions naturally become more sophisticated.

Early on, someone might ask:

What was revenue last quarter?

Then:

Which customers drove the decline?

Then:

Which of those customers are showing other signs of churn risk?

Eventually, a single business question may require five or six data retrievals, multiple calculations, data manipulation, and a final step that brings everything together into a useful result.

Supper already gives teams a semantic model that helps the agent understand their business and answer questions accurately.

But as the complexity of those workflows increased, another problem emerged.

Customers could get to the right answer, but they had to remember how they got there.

Before Skills, users might chain together a series of requests in Supper, pull intermediate results into spreadsheets, or try to recreate the workflow in a general-purpose AI tool.

Those approaches can work for one-off analysis.

They become much harder to manage when the process needs to happen again and again.

Because the real pain often isn't the first time you perform a complicated analysis.

It's the next time.

Once you've figured out how to get to the right endpoint, you shouldn't have to figure it out again.

From questions to reusable data workflows

Consider a customer success team building a customer health score.

There probably isn't one query that produces the answer.

The team may want to evaluate customers across several dimensions: usage, commercial activity, support signals, engagement, or other indicators specific to the business.

Producing the final health score could require several separate data retrievals followed by transformations, calculations, and reasoning to bring those signals together.

Without a reusable workflow, someone needs to reconstruct that process each time the analysis runs.

With a Skill, the process itself can become reusable.

The Skill can define what the agent should consider, what information it needs, which tools or calculations it should use, and what a successful result looks like.

The same idea applies to workflows like churn analysis.

A team trying to identify churned or at-risk customers may need several cuts of sales pipeline and customer data before arriving at the correct population.

Or consider risk scoring.

A Skill might guide the agent through retrieving a collection of customer attributes, calculating the relevant factors, and applying the appropriate scoring process.

These aren't simply questions.

They're analytical processes.

Skills give those processes a home.

Skills are the next layer of the Supper semantic model

Skills are particularly powerful because they aren't a separate automation system bolted onto Supper.

They are part of the Supper semantic model.

At the foundation of the semantic model are the structures of your data: tables, columns, relationships, and schema information.

Above those structures, Supper can capture descriptions and business context.

Then there are Terms, which allow teams to define reusable business concepts, calculations, and logic.

Skills sit at the top of that model.

One way to think about the progression is:

Tables and columns describe where the data lives.

Descriptions explain what the data represents.

Terms define what the business means.

Skills preserve how the business works with its data.

A Term might establish the definition of an active customer or a particular business calculation.

A Skill can combine those concepts into an entire process for analyzing customer health.

That distinction matters.

A semantic layer becomes far more useful when it can capture not only the vocabulary of an organization, but also its analytical processes.

A Skill is not just a saved prompt

Saved prompts are useful when you want to reuse an instruction.

Skills are designed for something broader.

A Skill can provide Supper's agent with the scaffolding necessary to accomplish a task while still allowing the agent to reason through the workflow.

For example, a Skill can define success criteria rather than prescribing every individual action.

It can tell the agent what a successful outcome should contain and allow the agent to work toward that result.

A Skill can also include reasoning loops when the task requires iteration.

If additional investigation is needed, the agent can continue working within the framework of the Skill rather than blindly replaying a fixed sequence.

And Skills can reference the wider Supper environment: other Skills, semantic-model context, algorithms, and tools.

That makes them particularly well suited to complex data workflows where the exact path may vary even when the desired outcome remains the same.

Skills work with the tools Supper already has

When you invoke a Skill, the Skill gives the Supper agent context and guidance for completing the task.

The agent then uses the tools already available to it.

That might include retrieving data, manipulating data, performing calculations, or working through other parts of an analytical process.

Supper can also combine different modes of data work as the task requires, including SQL-based retrieval and Python-based manipulation.

The Skill provides the framework.

The agent performs the work.

Some Skills can even bring specialized tools with them. Supper's chart creation capability, for example, can operate as a Skill with its own purpose-built toolset.

This architecture lets Skills support workflows that are much richer than a static sequence of instructions.

Skills remain governed by Supper

Making a workflow reusable shouldn't create a new path around your data controls.

Skills sit on top of Supper's existing platform and tools.

If completing a Skill requires data retrieval, the agent still retrieves that data through the underlying Supper tools. The user's existing entitlements and permissions therefore continue to apply.

A Skill doesn't grant access to data someone couldn't otherwise access.

It guides the agent in performing the workflow using the access and capabilities that user already has.

Skills are also part of the versioned semantic model.

As Skills are edited over time, those changes are logged, and users can inspect their lineage and move through previous versions.

That matters because workflows evolve.

The definition of a health score may change. A risk model may incorporate a new input. A team may discover a better way to perform an analysis.

The process can evolve without losing its history.

Build Skills manually or with an agent

Not every user should have to translate a complex analytical process into a formal workflow by hand.

Supper Skills can be created manually or with a dedicated Skills agent.

That means a workflow can begin the same way many useful processes begin: by working through the problem.

Once you have reached an outcome worth repeating, the process can be turned into a reusable Skill.

For more structured workflows, the Skill can specify required inputs.

For more open-ended workflows, it can give the agent broader goals and reasoning guidance.

When a Skill is invoked and additional information is required, the agent can work with the user to collect what it needs before continuing.

The bigger idea: your data system should learn the process

For years, semantic layers have focused primarily on defining the meaning of data.

That remains essential.

An AI agent cannot reliably work with business data if it doesn't understand the organization's metrics, concepts, relationships, and terminology.

But meaning is only one form of organizational knowledge.

There is also process knowledge.

How does your customer success team assess account health?

How does finance investigate a particular type of variance?

How does revenue operations identify a specific class of pipeline risk?

How does an analyst move from several raw signals to the artifact an executive actually wants?

These processes are often scattered across prompts, spreadsheets, documentation, dashboards, and individual employees' knowledge.

Skills create a way to preserve more of that process directly inside the data system.

The progression is straightforward:

Ask a question.

Develop an analysis.

Find a process that works.

Turn it into a Skill.

Use it again.

As AI data tools become more capable, the most valuable systems won't simply answer increasingly complicated questions.

They'll help organizations preserve what they learn about how to answer them.

That's what Supper Skills are built to do.

FAQ

What are Supper Skills?

Supper Skills are reusable frameworks that guide Supper's agent through complex, multi-step data workflows. They can capture inputs, instructions, success criteria, reasoning guidance, tools, algorithms, and relevant semantic-model context.

Are Supper Skills just saved prompts?

No. A saved prompt preserves an instruction. A Supper Skill can preserve an entire analytical workflow, including required inputs, reasoning loops, tools, calculations, and success criteria.

What are some examples of Supper Skills?

Skills can support workflows like customer health scoring, churn analysis, risk scoring, and other recurring analyses that require multiple data retrievals, transformations, calculations, or reasoning steps.

Where do Skills fit into the Supper semantic model?

Skills sit at the top of Supper's semantic model. Tables and columns define the data, Terms capture reusable business meaning and logic, and Skills capture reusable processes for working with that data.

Do Supper Skills respect existing data permissions?

Yes. Skills use Supper's underlying data tools, so the user's existing entitlements and permissions continue to apply when the agent retrieves or manipulates data.