I named one of my Hermes profiles “co-founder.” A little ambitious? Sure. But it describes what I want from it: help moving the work forward, rather than another conversation I have to keep restarting.
We discuss a roadmap together. That matters. The profile is proactive about items we’ve already talked through; it isn’t making up priorities out of thin air. It has meetings in the morning, mid-afternoon and at night, considers what needs doing, and acts on the work in that roadmap.
When something needs my review or approval, it pings me during the day. I still make those decisions. What changes is how much of the preparation I have to do before I can make them.
Five blog posts, ready for me to review
Here’s a concrete example. My co-founder profile handled blog-post creation for one of my clients. It brainstormed topics, then executed our blog content-writing skill bundle to create five posts.
The skill bundle already contained detailed instructions for the content. That’s why I felt comfortable giving it the heavy lifting: the writing brief was there, rather than a vague request to “write something good.” I considered the posts high-quality, but that’s my assessment of this assignment, not a benchmark of every Hermes setup.
It uploaded the posts to Google Docs and pinged me: “The five blog posts are ready for review. Here are the links.”
That’s the handoff I want. I open the documents and read them. If I see an issue, I explain what needs changing and ask it to fix it.
The sequence was straightforward:
- Brainstorm topics for the client assignment.
- Execute the detailed blog-writing skill bundle.
- Create the five posts and upload them to Google Docs.
- Send me the links for review.
- Revise if I identify issues.
Notice where the work ends: with drafts I can inspect. Preparing those documents is different from publishing to a client’s site. This example also doesn’t mean Google Docs access appears automatically in a new installation. Your setup needs the appropriate tools and access.
Give “proactive” something to work from
“Be proactive” sounds good, but on its own it leaves a lot unstated. About which project? What should happen first? What can the agent do without checking with you?
In my case, the roadmap gives that initiative a source. We discuss the work, and the profile acts on those discussed items. The point isn’t to hand over the job of deciding what the business should be. It’s to get help carrying agreed work forward.
If you’re starting out, I’d make the next deliverable very concrete. “Prepare a sourced comparison of these three options” gives you something to review. “Take care of the project” leaves the scope, result and permission boundaries fuzzy.
You can also separate preparation from commitment. Collecting information and drafting a recommendation are one set of tasks. Choosing a supplier, sending a message or spending money are additional actions. Write down which decisions you’re keeping. Those are suggested boundaries for your own setup, not a description of every permission in mine.
A rhythm that fits your work
My profile has those three meetings: morning, mid-afternoon and night. You don’t need to copy that frequency to get value from the idea.
If you want to try a similar routine, here’s an agenda I’d suggest. This is a starting point for you, rather than a claim about the exact agenda of my meetings:
- Morning: check the agreed priorities and identify the next useful deliverable.
- Mid-afternoon: surface blockers and questions that would stop progress.
- Night: summarize what changed and suggest what to carry into tomorrow.
One check-in might be plenty. More notifications aren’t automatically more useful. The question is whether the routine helps you see progress and make decisions without creating another inbox you have to babysit.
Make the review request useful
For me, the Google Docs links made the blog-post handoff easy to act on. I could go straight to the work instead of fishing five drafts out of a long chat.
I’d apply that same principle to other tasks. Ask the agent to name the roadmap item, link to the result, flag what’s unresolved and tell you what decision it needs. “The draft is ready; please review these two open questions” is easier to deal with than “Any thoughts?”
Decide what happens while it waits. Can it work on another independent draft, or does the affected task need to pause? If you request a change, make sure the next message points to the revised document. Otherwise you can end up reviewing an old version without realizing it.
Silence shouldn’t quietly become approval. Make that expectation explicit in your instructions and verify that your setup follows it.
Remembering the roadmap is different from scheduling work
Hermes has documented persistent memory. That can help retain context, but memory, scheduling and permission are separate pieces.
Remembering a roadmap doesn’t itself schedule a meeting. A scheduled meeting doesn’t itself authorize every action the agent proposes. When you configure your own routine, check each piece: the roadmap it uses, when it runs, where it sends the result and how you stop the recurrence.
Use the documentation for your installed version. A good prompt can define the job; it can’t stand in for a tool or connection that hasn’t been configured.
Start with one result you can inspect
My suggestion for a first attempt is modest: one public-source brief, one checklist or one draft in a test area. Tell the agent what to produce, which sources to use and where to put the result. Keep private accounts and consequential actions out of that first exercise.
Then inspect what it actually made. Does it answer the brief? Are missing facts visible? Does the notification take you to the right draft? Record any corrections before adding more tasks or recurring check-ins.
That exercise is a suggestion, not a test I’m claiming to have performed for this article. My client-blog example is my own reported experience; the client documents and exact configuration aren’t reproduced here, and this isn’t a matched product test.
If you want the product details, read the Hermes profile. If you’re weighing alternatives, use the Hermes–OpenClaw comparison. For a first setup, start with the beginner guide.
For me, the useful part is simple: the heavy lifting arrives as work I can review. I still have to read it, decide and give feedback. But I’m starting with five documents, not a blank page.
