Claude Code Process · AI AUTOMATION INSIDERS

Claude Code Process

I'm building out my business processes using Claude code. Wondering if anyone's got advice or can share their own process? Would be awesome to hear how people are effectively leveraging "run background tasks" for simultaneous tasks, /skills and /agents to get high quality results.

My current process is:

Use claude code plan mode to develop a detailed step by step walkthrough for AI to complete the task (i.e. scrape instagram posts for trends), I specifically reference any files or functions to be used with the @ designator
Walkthrough always includes instructions to clarify with human in the early stages / check outputs (i.e. check 10 API returns for suitability before starting bulk scraping) & update the file if there are any errors / improve for next time
Start a new claude chat, reference the walkthrough file and run it using Haiku

I think the benefits of this approach are:

Can have human interaction (which I don't get with /skills or /agents) at least until I can verify performance >95%
Saves on credits by using Haiku once detailed instructions or python functions are already developed
If I want to run another task simultaneously I just boot another terminal with a fresh claude session

FYSA my setup is; Windows, VS Code, WSL Terminal (I would go crazy with powershell). Always use python for functions. Supabase for databasing. Mix of .md and .yaml files for unstructured data. Planning to add cron jobs next...

0 comments