List Skill Types
list_skill_types returns the complete active skill catalog used by Agent Studio. Use it before recommending skills for an agent. The category is Agent Operations.
Inputs and results
Call the tool with an empty JSON object:
{}
The runtime supplies the authenticated organization context. Do not supply an organization ID or other properties. Invalid arguments, missing organization context, and catalog failures return an explicit failure.
The result contains success, message, and skillTypes. Each skill type includes name, apiName, label, functionName, description, category, and isActive.
Results preserve registry order and canonical names. All active entries are returned, without filters, paging, or truncation. Inactive entries are excluded. Descriptions are catalog summaries; full instructions and tool parameter schemas are not included.
Use with AgentBuilder
- Read the catalog before recommending additions. Reuse a successful result within the conversation.
- Compare candidate skills with the target agent's assigned skills and business objective.
- Recommend the minimum useful set. Explain each proposed addition and its purpose.
- Obtain approval through the existing AgentBuilder approval flow. Apply approved configuration with
dml_sobjects, then verify the saved records.
Agent Studio administrators select from this same active catalog when they add a skill. Discovery does not assign skills, load instructions, append tools to a request, or permit execution of the listed tools. A recommendation alone makes no configuration changes.
Activation
The release operator must update AIHandler. Update JobQueueHandler or WebAPI before using this skill through their background or public-chat paths.
In the authoring AgentBuilder, add List Skill Types, set it active, and use Run Context All. Apply the revised agents/AgentBuilder.md instructions and publish a new package. The release operator must set the system-agent registry to the actual returned package ID and activate that release. Start a fresh Builder conversation after activation. Editing the Markdown source alone does not activate the change.
Deployment and publishing are operator-controlled. This extension requires no Salesforce metadata deployment.
Future direction
Admin-approved tool sets with deferred loading are a separate future requirement. Catalog discovery, permission to execute, and loading tool definitions must remain separate. This release does not provide automatic tool loading or self-assignment.