I have a broad but shallow technical background (longtime management consultant) and use agents (a variety of services and models in VS Code) extensively for my own projects. I know I'm hitting a ceiling. My workflow is naive and nonpro, a lot of tinkering and bashing my way through code generation, context drift, manual fixes, etc.

I want to learn how senior engineers treat these tools. Processes, systems, guardrails, must-dos, etc.

If you write code for a living and have integrated agents into your workflow, what does your system look like?

Any/all input is welcomed: resources, URLs, outlines, warnings, discoveries, recommendations.

None of these are specifically needed, but ideas?

-How do you structure a brand new project? Scaffolding, git init/ignore? Repos? Initial commit strategies? How do you keep stuff out of the context window that you don't want in it?

-How do you "layer" your work so it's much more about the context and integrity of the structure and much less about prompts?

-Do you switch models (often? ever)? Why?

-What other tooling do you have alongside basic agents/environments?

-What are "pro habits" that I don't have and should?

-Any suggestions for how to work in a very domain-specific endeavor to try to increase the utility of the agents for a silo like this?

> How do you structure a brand new project? Scaffolding, git init/ignore? Repos? Initial commit strategies?

Don't overthink it. Use whatever bootstrapping tools your project framework comes with (rails app? use `rails new`)

> How do you keep stuff out of the context window that you don't want in it?

Don't overthink it. Just use one chat per general topic. New topic -> new chat. Save anything relevant to the project that the agent doesn't pick up in new chats in AGENTS.md

> How do you "layer" your work so it's much more about the context and integrity of the structure

Save project specific needs/requirements in AGENTS.md or docs/ or skills or whatever and tinker until it works

> Do you switch models

Nowadays just for code review. Recently gpt5.5 has felt very good for general dev

> What other tooling do you have alongside basic agents/environments?

Neovim for code browsing (don't write much code anymore but the code editor is still useful for reading code) and various plugins/tools i've built up over the years

Honestly, there's no blanket solution. The best engineers I've worked with all have different tools custom to their workflows

> know I'm hitting a ceiling. My workflow is naive and nonpro, a lot of tinkering and bashing my way through code generation, context drift, manual fixes, etc.

Hit a ceiling? Fix it. Typically it's

- do work

- notice something annoying about your workflow

- fix it systemically - research different solutions - sometimes it's writing a small script (agents are great for this), sometimes it's finding a new tool, sometimes it's overhauling your entire system, it just depends

- do work

repeat...

there's really no one stop solution for a perfect workflow

I don't find that the models are so capable that the workflow can be "much less about prompts". you still need to be in the loop, prompting and validating, to get good output. people will say otherwise -- I say those people are full of shit, have an adverse incentive, or don't have high standards.