the problem
i’ve lost count of how many times i’ve sat down to build something and spent the entire session just trying to figure out what to build. not the code — the plan. what’s the stack? what’s in scope? what’s out of scope? what does done even look like?
by the time i’d answered half of those questions, i’d burned through my motivation and closed the laptop.
that’s analysis paralysis. and if you’re a solo builder working with ai tools, it’s the silent killer. the ai is ready to work — but if you can’t tell it what to build, nothing happens.
what project-planner does
project-planner is a repo you open in claude code, codex, or gemini cli. there’s no code in it. no dependencies. just a set of skill files that guide the ai through a structured interview with you.
it asks one question at a time. it starts with the problem, not the solution. it makes you define success criteria and out-of-scope before you ever talk about technology. and when the interview is done, it produces a clean /outputs/ folder with four files:
- PRD.md — what you’re building and why
- STACK.md — tech decisions with rationale
- TASKS.md — a realistic first sprint
- context file — full project context for whatever ai tool you use next
you drop that folder into your build repo and the ai has everything it needs from the first message. no re-explaining.
why i think this matters
most people don’t need help writing code anymore — the ai tools are genuinely good at that part. what they need help with is the part before the code. the thinking. the scoping. the “what am i actually trying to do here?”
project-planner forces that conversation to happen in a structured way. you can’t skip out-of-scope. you can’t jump to the stack before the problem is locked. the skill files won’t let you.
for someone like me — not a traditional developer, working solo, building tools for my own workflows — this is the difference between a productive session and a wasted afternoon.
how the chain works
project-planner is the first link in a two-repo chain:
project-planner → outputs/ → mvp-builder → ship
plan in one repo, build in the next. each repo has one job. simple.
try it
it’s a github template. clone it or hit “use this template,” open it in your ai tool, and say /init.