Understanding SuiteCloud Agent Skills
SuiteCloud Skills are reusable packages that provide task-focused AI capabilities to help you automatically complete tasks in NetSuite. Skills use generative AI to interpret user intent and produce helpful outputs with structured content.
Skills can help you:
-
Standardize how common tasks are executed, such as reviews, validations, or documentation drafts.
-
Reduce errors by encouraging the use of source-of-truth references.
-
Improve security and compliance by embedding required practices, such as least privilege, avoiding Administrator access by default.
-
Capture best practices in a reusable format.
How SuiteCloud Skills Work
Skills are defined in a markdown file that explains what it does, when to use it, what inputs it needs, what outputs it produces, and what rules it follows. The generative AI tool uses the instructions and executes in the skill file to guide how it responds to a prompt.
When a skill is invoked:
-
The AI interprets the user request.
-
The skill instructions define how the task should be handled.
-
Reference sources are used to validate information.
-
The AI returns output in a structured and consistent format.
You should always review generated output and test any changes in a non-production environment before deployment.
A skill file can include the following:
-
Purpose and scope - What the skill is designed to do and what it should not do.
-
When to use it - The scenarios where the skill applies.
-
Inputs - The information the skill expects (for example, an SDF XML snippet, a role name, a business requirement, or a record type).
-
Workflow - The steps the skill should follow to complete the task consistently.
-
Decision rules and guardrails - Constraints the skill must follow (for example, least privilege, don't guess IDs, validate against a source of truth).
-
References - Source-of-truth files or documentation the skill should rely on to produce accurate results.
-
Output requirements - What a complete answer must include (for example, exact IDs, recommended levels, and a short justification)/