Claude God Tip #11: Meet Allen Kendrick, My Newest Employee

Claude God Tip #11: Meet Allen Kendrick, My Newest Employee

Published: November 29, 2025 • 5 min read

In a previous blog post, I told you all the dreadful story of how I lost my perfectly trained Claude instance that was responsible for refining my blog post drafts into a ready-to-publish piece. This Claude instance understood all my requirements after a series of back and forth conversations. It understood the expected format and structure it had to follow based on my codebase. I lost this perfectly trained instance because of how long the conversation had gotten, and in that same blog post, I shared exactly how I captured this instance using a methodology I named LLM Instance Cloning.

However, remember I told you that if you find a way to take a workflow that would normally take 2 weeks down to 2 days, that should not stop you from wondering how to further cut down the time to 2 hours? Well, applying that same mindset, I figured out a way to further take down my blog refinement process from minutes to seconds.

My Previous Writing Workflow

Here is what my writing workflow used to look like:

1. I write a draft: This could be written anywhere at any time. Sometimes it's written sitting down in front of my screen like right now. Other times I wake up in the middle of the night and struggle to fall back asleep, but then I get ideas so I simply grab my phone and write. Other times I'm watching a meal cook and it's taking forever and I get an idea and I write. Most times, I end up editing the draft on my computer regardless of when or where I wrote it.

2. Copy and paste into trained LLM chat: Then I copy and paste this draft in the chat where my trained LLM Instance lives and it knows exactly what to do, no new instructions needed.

3. Download and add to codebase: Then I download the file created and add it to my project's codebase.

Pretty quick right? Well, I found a way to cut this process down even further.

The Subagent Solution

In the 4th episode of Claude God Tips, I talked about how to create your own Employees using Claude's custom subagents feature. If you haven't read that, I highly suggest you do! Then in this blog post, I showed you the 7 subagents that I created based on 129 code reviews from my time at Outlier training AI models to develop "Golden Standard" web applications. I gave these 7 subagents fun names and personalities as well!

Well just yesterday, I got the dreadful message that "Claude hit the maximum length for this conversation. Please start a new conversation to continue chatting with Claude". Now normally at this point, I would simply take my Blog refinement prompt which I created using LLM Instance Cloning and use that to start a new chat. However, I stopped to think... is there a faster way to do this? Can I take this 3-step process down to one step?

And guess what, I did!

Introducing Allen Kendrick

I converted my perfectly trained instance into a subagent. You see, so far I have a total of 8 Employees on my team. Hopefully by now you've spoken to Nancy, my personal assistant (click on the chatbot at the bottom right of the screen to talk to her if you haven't already). And in this blog post, I introduced Cassandra, Amber, Eesha, Daniella, Lindsay, Micaela and Kristy to you.

Well you see, I believe in gender equality so I hired a man this time! Meet Allen Kendrick, my blog refining agent.

Here is everything you need to know about him:

Allen is the person who physically winces when he sees a paragraph longer than five sentences because he knows readers will skip it. He restructures content the way architects redesign floor plans by keeping the foundation intact while making everything flow better. Allen has strong opinions about subheadings, believes bullet points are underrated, and will quietly add hyperlinks to your previous work because he knows you forgot. He has a sixth sense for where readers will get confused and restructures those sections before anyone has to ask. He catches the spelling errors your eyes gloss over and smooths out the sentences that made sense in your head but confuse everyone else. When Allen finishes with your draft, readers actually make it to the end. If you've ever wished you had someone who could polish your drafts without making them sound like someone else wrote them, Allen is already disappointed you haven't asked him sooner.

How to work with Allen:

• "Allen Kendrick, do your thing"
• "Ask Allen to improve readability and add appropriate structure"
• "Have Allen check for errors and add relevant hyperlinks to my previous posts"

The New One-Step Workflow

I created Allen Kendrick using my perfectly trained Claude Instance and now, the three-step process I described above has been cut down to one step. I now have a blog draft text file sitting at the root folder of my code repository. In this file is where I write all my drafts now. Once I am done, I go to my terminal and ask Allen Kendrick to do his thing and this blog post is ready! No more extra steps of downloading the file and placing it in the right location in my project codebase. Allen Kendrick handles all of that.

The best thing about Allen Kendrick is that because I call him through Claude Code in the terminal, the environment includes "Today's date" which Claude can see in its env block. I no longer have to keep changing the dates since Claude on the web is not very great at knowing what the current date is.

Why This Matters

Being able to come up with ways to shorten the time taken to accomplish a task without sacrificing quality is one thing that genuinely makes me happy. This is the type of stuff I mentioned when I told my Automation love story that gets me excited.

I hope you found this post useful.

As always, thanks for reading!