How to Automate Work with Six Hermes Agent Uses

TL;DR

Hermes Agent can handle long-running projects, manage tasks through a Kanban board, research competing apps, preserve work in a memory wiki, assist with computer administration, and plan around a daily priority. Detailed prompts, clear constraints, and clarification questions make these workflows more useful, while the agent’s browser and computer controls let it perform research alongside other ongoing tasks.

Transcript

Hermes Agent is the most powerful AI software out there right now. The issue is 99% of people have no idea what to do with it. In this video, I'm going to go over six use cases that have totally transformed my workflow. By the end of this video, if you stick with me, you'll have a 24/7 AI employee doing work for you around the clock. Now, let's loc... Read More

Key Insights

• /goal is designed for long-running, complex tasks rather than ordinary one-step requests. The creator reports that a /goal assignment ran for more than 24 hours, making the feature suitable for detailed applications, extensive modifications, and long documents when it receives sufficient direction.
• Metaprompting is the key to getting useful results from /goal. Instead of submitting a vague request such as building an app, the user first asks an AI to construct a detailed /goal prompt that captures requirements, constraints, desired behavior, and relevant project details.
• Clarification questions can prevent hours of misdirected autonomous work. In the demonstrated game project, Hermes reviews the detailed prompt and asks several questions before beginning, allowing important uncertainties to be resolved before the agent commits substantial time to implementation.
• The Hermes Kanban board automatically distributes tasks placed in triage to agents and subagents. It is opened by running "Hermes dashboard" in a terminal, following the supplied URL, and navigating to the Kanban area where assignments can be created and monitored.
• A daily Kanban routine separates agent-compatible assignments from work that still requires the user. Example delegated items include scripting a Hermes video, drafting three tweets, and researching current AI tools, while paying a credit card bill is identified as something Hermes cannot do at that moment.
• Hermes can research competing applications by controlling a browser and computer. It can navigate a site, click through pages, inspect the console, collect images, open a terminal, create its own tasks, and request permission before searching for tools and libraries.
• A technical competitor report can document an application’s high-level design, technology, features, payments, prices, analytics, and collected events. The report can be saved as a Markdown file and passed to coding agents so they can incorporate relevant ideas into another application.
• A memory wiki functions as an automated diary and a reinforcement source for the agent’s memory. It organizes subjects, conversations, completed work, and daily logs into clickable pages, helping both the user and Hermes revisit earlier activities and contextual information.
• More videos from this channel:
  • How to Use Omarchy for an AI-First Workflow
  • How to Use Claude Fable 5.1 Without Overspending
  • Is Claude Opus 5 Better Than Claude Fable 5?
• More from this channel: Alex Finn

Questions & Answers

Q: How do you use /goal effectively in Hermes Agent?

Use /goal for work that genuinely benefits from extended autonomous effort, including complex applications, detailed changes, and long documents. Before launching it, ask an AI to create a comprehensive /goal prompt from your project description. Include concrete requirements and constraints, then let Hermes ask clarification questions before it begins. A vague instruction such as "build me an app" will behave more like an ordinary request and will not unlock the feature’s full value.

Q: What is metaprompting for a Hermes /goal task?

Metaprompting means asking an AI to write the detailed prompt that Hermes will execute. Describe the intended project, provide its important details, and explicitly request a /goal prompt. The resulting prompt can be much more extensive and opinionated than one written quickly by hand. In the example, this process turns a description of a 3D third-person shooter made with Godot into a structured prompt suitable for hours of autonomous work.

Q: When should you use /goal instead of a normal prompt?

Use /goal when the assignment involves sustained, complex work, such as developing a detailed application, implementing extensive changes, or writing a long document. The creator reports using it for runs lasting hours and, in one case, more than 24 hours. A normal prompt is more appropriate for a straightforward request. Even with /goal, the task needs detailed requirements and constraints or Hermes may simply produce an ordinary result.

Q: How does the Hermes Kanban board automate a daily workflow?

The Kanban board accepts assignments in a triage column, automatically assigns them to Hermes agents or subagents, and moves them toward active work. The recommended routine is to begin the morning with a written to-do list, identify everything Hermes can handle, and create those items in triage. While the agents complete scripts, tweets, research, or reports, the user can focus on responsibilities that still require direct human action.

Q: How do you open and use the Hermes Kanban board?

Open a terminal, enter "Hermes dashboard," and follow the URL it provides. Navigate to the Kanban area and create assignments in the triage column. Example tasks include scripting a YouTube video about Hermes use cases, producing three related tweets, and researching current AI tools for a report. Hermes then assigns the work to agents and moves it from triage into the workflow without requiring manual assignment.

Q: How can Hermes Agent analyze a competing application?

Ask Hermes to open a browser, visit the competing application, and produce a full research and technical breakdown. The request can ask which technology stack the site uses, how it works, which features it offers, and what ideas could be emulated. Hermes can navigate the site, inspect the console, gather images, use a terminal, identify tools and libraries, and compile its findings while the user continues other work.

Q: What information can a Hermes competitor report contain?

The demonstrated report includes a high-level overview, the technology behind the application, its payment system, prices, analytics, and the events being collected. Hermes obtains the information by navigating the site and examining relevant technical details through its browser and computer controls. The user can request that the finished research be saved in a Markdown file and then provide that file to another coding agent working on an application.

Q: How do you create and use a Hermes memory wiki?

Ask Hermes to build a site containing a list of subjects discussed together and daily logs of completed work. Each subject or log should be clickable so the user can review additional details about earlier conversations and activities. The resulting wiki works as an automated diary for the user and reinforces Hermes’s memory because the agent can revisit those records when it needs context about previous work or a familiar subject.

Summary & Key Takeaways

• Use /goal for complex, long-running work such as building detailed applications, making substantial changes, or drafting long documents. Before starting, ask an AI to create a detailed, opinionated prompt containing the project requirements and constraints. Hermes can then ask clarification questions and continue working for hours, with one cited run exceeding 24 hours.

• Begin the day by writing a personal to-do list and transferring suitable assignments into the Hermes Kanban board’s triage column. Hermes automatically assigns those items to agents or subagents and moves them into active work. This lets the user complete tasks requiring human action while agents handle scripts, social posts, research, and reports.

• Hermes can inspect competing applications through its browser and computer controls, producing technical reports on their stacks, features, pricing, payments, analytics, and behavior. A memory wiki can separately organize discussed subjects and daily logs into a searchable personal record. A scheduled morning prompt can connect daily priorities with agent-generated tasks and updated memories.