Craig Hewitt, founder of Castos, demonstrated on Zero to Agent in 30 Minutes how to convert a fresh Hermes installation into an always-on social media agent that can learn a person’s writing, draft posts, and run recurring research tasks. His walkthrough emphasized the context, workflows, and safeguards needed for an agent to produce reliable, useful output.
Key points from the demonstration
The demo outlined several practical steps and recommendations:
-
Choose the right agent setup: decide whether you need an interactive tool for hands-on work or an always-on agent that runs on a schedule. Craig used the Hermes desktop app in the demo; that app can later be deployed to a cloud server or a dedicated machine.
-
Create a structured workspace: split the project into separate files for voice guidance, editorial standards, post templates, examples, and operating instructions so the agent can retrieve consistent guidance as it works.
-
Seed the agent with your own context: provide examples of your posts, emails, and other writing that reflect the desired style. Craig also added writing he likes from people he follows to broaden the agent’s reference set.
-
Turn examples into a voice system: have the agent analyze the material and document its findings. The voice profile captures audience, point of view, sentence style, recurring themes, editorial rules, and typical post types.
-
Test a narrow workflow with human review: start with a single task (for example, drafting several LinkedIn posts from a supplied idea) and keep a human in the loop to evaluate output, correct mistakes, and refine instructions.
-
Package repeatable work into skills: create reusable procedures for recurring tasks such as researching topics, selecting a post format, retrieving relevant examples, and drafting in the approved voice. Craig likened these skills to standard operating procedures that make repeatable work more consistent.
-
Connect the agent to fresh data: add sources of new ideas — news feeds, websites, social platforms, or internal systems. Craig recommended beginning with a simple, semi-automated trend scan before building a more complex data pipeline.
-
Add triggers and safeguards: decide what starts each workflow (schedule, user request, webhook, or change in another system). Use separate accounts and limited permissions for autonomous agents so you can trace actions and control access.
When does an agent become useful?
Craig emphasized that agents become valuable when they have adequate context, clear processes, the right tools, and enough oversight to validate each workflow. With those elements in place, teams can move from ad-hoc prompting toward systems that monitor information and complete recurring work autonomously.
What’s next
Next week’s episode will feature Max Johnson, cofounder of briix.ai, who will rebuild a workflow that currently lives in someone’s head (or a messy Notion doc or long email thread) into an autonomous agent, live from scratch. Viewers will be able to follow every decision: how to identify steps that can be handed off, how to handle steps that can’t, and how to structure the overall system so it runs without human intervention.
Two upcoming training sessions were also announced: "Harness Engineering for AI Agents" on August 12, focused on designing and building production-ready agent infrastructure, and "Build Your First Local Agent with Hermes" on August 26 for deeper hands-on Hermes experience.
