Create reusable AI skills in ChatGPT and Claude
A short Skills 101 walkthrough of reusable AI workflows, with email triage as the example.
01From a repeated prompt to a repeatable task
Repeated work needs more than a prompt you keep pasting into a chat. In this walkthrough, James Bell introduces a way to turn a familiar task into a reusable workflow. The accompanying guide explains how to define the trigger, inputs, rules and finished output before you build.
02Three things to take into your own workflow
- Define the job clearly. State when the skill should run, what information it receives, what it should do and what the result should look like. These four questions make the workflow easier to explain, test and reuse.
- Make the output actionable. The email example sorts messages by priority and records the subject, sender, summary, action owner, required response and deadline. If an action or deadline is missing, the workflow should say so.
- Test the rules on different inputs. Check ambiguous messages, overlapping priorities and cases needing judgement. Tighten instructions or add examples where the standard does not hold.
This video records the workflow demonstrated in July 2026. Installation, connected accounts and tool access depend on the product, plan and workspace permissions. A reusable instruction set does not itself grant access to an inbox.
Read the full skills guide, or explore AI Operating Partner for help embedding repeatable workflows in your business. For investment teams building capability, see practical AI training.