Agentic Workflow Experiments: Tech Debt Epic and Local Agents

TL;DR

• Agent 6451 posted a Slack message suggesting we pull items from the AI tech debt epic as candidates for testing new agentic workflows
• Used Open Code to assess everything in that epic and generate a list of one-shot implementations
• Got a list with top four items, next three, and a list not good for one-shots
• Plan is to develop my current local flow and get it running using flows and Agents through the UI, modeling that today
• Also trying to teach my local agents to never use specific words and phrasing

Testing New Agentic Workflows

Agent 6451 posted an interesting Slack message, essentially suggesting that items from the AI tech debt epic would make good candidates for testing out new agentic workflows. That’s exactly what I’d like to do. Using Open Code, I asked it to assess everything in that epic and provide a list of one-shot implementations. What I got back was a list of several items: the top four, then the next three, followed by a list of items that aren’t good candidates for one-shots. My plan now is to develop my current local flow and get it running using flows and Agents through the UI. I’m going to model that out today and see if I can make it happen.

Agent Language Training

I’m also working on teaching my local agents to never use specific words and phrasing.