AI Agent Action in Workflows Adds Conditional Skills Library

TL;DR

HighLevel added a Skills section to Workflow AI Agent actions that loads reusable instructions conditionally. If you build AI workflows, you can now modularize your prompts.

Who this affects

HighLevel has not confirmed which specific plans or sub-account types have access to this feature. Users running or building automations with the AI Agent action can check their workflow builder directly.

What changed

• Added a Skills section directly below Tools within the AI Agent Action in Workflows.
• Introduced three ways to add skills: Create with AI, Write skill instructions, or Add existing skill from the library.
• Included four built-in skills in every account: Humanizer, Write-well, Cold-email, and Sms-writer.
• Configured the AI Agent to check skill trigger conditions on each execution and load full instructions only when those conditions match.
• Enabled multiple matching skills to load simultaneously within a single action run.
• Added shared skill management across an account, allowing skills to be edited globally, detached on a per-action basis, or deleted from the library.
• Limits, plan gating, and rollout status have not been confirmed by HighLevel.

Where to find it

Go to Automation > Workflows. Open a workflow, select or add an AI Agent action, and locate the Skills section positioned directly below Tools.

How agencies will use this

An agency setting up a lead response workflow can use the AI Agent action without hardcoding every style requirement into a single massive prompt. Instead, the agency attaches standard skills from the library, such as Cold-email or Sms-writer, alongside custom-written skills.

When the workflow executes, the AI Agent checks the trigger conditions, dynamically loads the relevant skill instructions if the conditions match, and runs the action. If the agency needs to tweak communication standards later, they can edit the skill globally across the account library rather than updating every individual workflow action manually.

Our take

This is a major architectural improvement to the AI Agent action. Monolithic prompts that try to handle every scenario are hard to maintain and prone to drifting. Loading instructions conditionally and managing them in a shared account library brings standard software-style modularity to workflows. If your sub-accounts rely on AI Agent steps, start moving common instructions into skills.