AI Agent News/stories/databricks-explains-how-to-build-reusable-skills-for-genie-one.md

Frameworks

Databricks explains how to build reusable skills for Genie One

Databricks has published a guide to building skills in Genie One, reusable workflow recipes that capture a team's steps, checks and output formats. The guide covers mapping a workflow, building a skill through conversation, testing it and publishing it through Unity Gateway Skills (Beta).

Source: www.databricks.com/blog/teach-genie-one-your-workflows-ultimate-guide-skills-genie-one

Genie One runs a skill on demand when a user calls it, or when it decides the skill is relevant to a request. Each skill is saved as a folder with a SKILL.md file in the open Agent Skills format, and a skill published through Unity Gateway can be used by any agent that supports MCP. Skills sit alongside Genie Ontology, which holds business definitions, and Genie Agents, which hold domain knowledge for one area.

Before building a skill, the guide recommends mapping the workflow: when it applies, which tools and data it uses, the order of steps, the checks that run every time and the expected output. Genie One then interviews the user about the method before it writes the skill. The guide advises strict rules for anything other teams or systems rely on, such as fixed formats and sign-offs, and leaving other steps to Genie One's judgement.

Testing follows the same approach, with a golden reference saved as the example to match, corrections made in conversation and a retest in a fresh chat. Skills can be shared with individuals or groups, and a scheduled task can run a skill on a regular cycle, with results sent to a chat thread and by email.

Why it matters if you build agents

Teams building agents can package recurring procedures as shared, versioned skills instead of repeating the same instructions in each chat. The guide suggests keeping skills small and chaining them, so an update to one shared skill reaches every skill that uses it. Review the generated instructions before relying on them, because each line is followed on every run.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe