The page has no enablement switch of its own. Agentforce has to be on first, and Agentforce Assets is simply where the reusable building blocks live afterwards.
- Open the Agentforce Assets page
From Setup, enter Agentforce Assets in the Quick Find box and select it. It opens on the Subagents tab, with Actions beside it.
- Read the standard assets first
Open whatever sits closest to your use case and check its classification description, scope, instructions and attached actions. Half the custom work teams plan turns out to be redundant.
- Create the custom asset in the library
Build it here rather than inside one agent when two or more agents will need it. Write the action description for the reasoning engine, because that text decides whether the action gets invoked at all.
- Add it to an agent, then preview the routing
Adding the asset in Agentforce Builder copies it immediately, so treat that click as a snapshot. Preview real utterances and confirm your subagent wins over any standard one covering similar ground.
Global on the Agentforce Assets page and addable anywhere, or local to one agent in Agentforce Builder.
A standard Salesforce asset for common patterns, or a custom one when the standard cannot reach your data.
A flow for declarative work, Apex where a flow cannot express it, or a prompt template for generative output.
- Editing a library asset feels like it should update the agents using it. It does not, and nothing warns you the two have drifted.
- Older docs say Topic where the interface says Subagent, and the metadata kept the old word in localTopics and pluginType.
- A vague action description produces a well-built action the agent never calls. Describe the trigger conditions, not the implementation.