@tmuxvim People are just behind, or they are gatekeeping the concepts out of fear. We should be helping as many people learn these techniques as we can.
@asaio87 I’ve been using the term pipeline myself for a few months now, before the community coalesced on the loop name. Even called my parent skill /pipeline, which kicks off the entire process.
Before you start looping any agent work, it’s imperative you learn the basics of test driven development, and how to focus the scope of your AI development to only getting the failing tests passing.
With loops, you can organize your project features into singular, digestible jobs for you AI agents to implement sequentially. With a properly configured Skills folder, automating your project with AI becomes a breeze.
@MarshallJGould Yes if you’re prompt is consistent and the clarifying questions are triggered regularly, there is memory or instructions in the Claude.md or elsewhere that the prompt is pointing
@beffjezos Imagine having an entire agency team on your terminal, and not using it to go into business for yourself while mainstream retail still resembles the old environment.
@DanielSmidstrup Opus 4.7, still too nervous to turn 4.8 back on after it started hallucinating bash output during 6+ worktree sessions two weekends ago. Bug case not resolved yet either
@diegocabezas01 TLDR: Create a skill that specs a multi-phase implementation plan. Create a test driven pipeline skill that is scoped to one phase section at a time. I give away my whole plan book that I use on my paying clients on my website, and I’m releasing a YouTube tutorial this week.