Agent Skills

Understand built-in agent skills, configure skill parameters, and create custom skill prompts to extend agent capabilities for your specific analytical needs.

What Are Agent Skills?

Agent skills are modular capabilities that give agents access to specific tools and behaviors. Each skill encapsulates a set of functions — such as querying data, fitting a statistical model, or generating a visualization — along with the logic for when and how to use them. Skills are the building blocks from which agent capabilities are composed.

An agent's skill set determines what it can do. The Data Analyst agent, for example, has skills for data querying, statistical analysis, and chart generation but not for writing formatted narrative reports — that skill belongs to the Report Writer agent.

Built-In Skills

Skill	Description	Default Agents
Data Cleaning	Detects and resolves nulls, duplicates, type mismatches, and encoding errors in datasets.	Data Cleaner
Statistical Analysis	Runs descriptive statistics, hypothesis tests, correlation analysis, and regression modeling.	Data Analyst, Anomaly Detector
Visualization	Selects appropriate chart types and generates chart configurations from data and a stated goal.	Data Analyst, Report Writer
Forecasting	Builds time-series forecasts using ARIMA, Prophet, exponential smoothing, and ML-based models.	Forecaster
Reporting	Generates structured narrative reports with executive summaries, data commentary, and embedded charts.	Report Writer
SQL Generation	Translates natural language questions into SQL queries optimized for the connected data source.	SQL Expert, Data Analyst
Anomaly Detection	Identifies statistical outliers and anomalous time-series patterns using IQR, Z-score, and ML isolation forests.	Anomaly Detector

Skill Parameters

Each skill exposes a set of configurable parameters that control its behavior. Parameters are set per agent in the agent configuration panel. Common parameters include:

Confidence threshold — For anomaly detection and forecasting skills, the minimum confidence level required before flagging an anomaly or returning a forecast.
Forecast horizon — For the forecasting skill, the number of periods to project forward.
Max rows — The maximum number of rows a skill will process in a single call, to prevent accidental large scans.
Output format — Whether to return results as a chart, a table, a narrative summary, or raw JSON.

// Example skill parameter configuration (JSON)
{
  "skill": "forecasting",
  "parameters": {
    "horizon": 90,
    "seasonality": "weekly",
    "confidence_threshold": 0.8,
    "output_format": "chart_and_summary"
  }
}

Creating Custom Skill Prompts

Custom skill prompts extend the behavior of a built-in skill with domain-specific instructions. For example, you can add a custom prompt to the Reporting skill that instructs it to always frame insights in the context of a specific business objective, use company-specific terminology, or apply a particular narrative structure.

Open an agent in the agent editor.
Click on a skill in the skills panel to expand its settings.
In the Custom Prompt field, enter additional instructions that will be appended to the skill's base system prompt.
Use {{metric_name}} and {{data_source}} placeholder variables if you want the prompt to reference the current query context dynamically.
Save the agent to apply the custom prompt.

Skill Marketplace

The skill marketplace (accessible from Settings > Agent Skills > Marketplace) provides community-contributed skill bundles that can be imported into your workspace. Each marketplace entry includes a description, the skill parameters it exposes, and a usage example. Marketplace skills are reviewed by the OpenAnalyst team before listing.

Note: Custom marketplace skills run within the same security sandbox as built-in skills. They can query your data sources only within the permissions already granted to the agent they are attached to.