How to use

Why: give your idea enough context

Imagine a service for a neighborhood book swap. Readers list books they can lend, find a book nearby, and arrange a handoff with its owner. Even a detailed feature request can leave the loan period, budget, and completion criteria undecided. If AI treats an unknown loan period as settled, you may review a plan built around a rule you never chose.

Write the actions and constraints you already know. kickoff combines that description with requests to ask about gaps, explain choices, agree on a plan, and verify work. You can inspect the actual generated example on the homepage. The aim is a shared plan you can review before development starts.

How: move from an idea to an agreed plan

Create and share a prompt

1. Have an external AI tool ready separately; kickoff does not provide one. Open Create a prompt on kickoff. Enter a service description: “Neighbors list books, request loans, and arrange a pickup.” This is the only required field. The name is optional; a blank becomes “Undecided”.
2. Add considerations such as budget, schedule, or a preferred example. Leave unknowns blank: they become “Needs clarification”, not “no constraints”. The input stays on this page and is lost when you leave or reload.
3. Read the preview: it is a request for your external AI tool. Copy it and paste it into that tool yourself. The prompt requests the startup guidelines and AI behavior rules, the only two required site documents, and reuse of your inputs. If copying fails, select the preview and copy manually. With JavaScript off, append the three fields to the base prompt yourself, then paste the request into your external AI tool.
4. Ask AI to confirm it has read both documents. If a URL cannot be opened, use the Markdown or AI behavior rules link on the prompt page and paste the missing text into the conversation. Continue with service details; do not assume unread rules were applied. Optional catalog access failure alone does not stop planning.

Create a custom prompt →

Agree on a plan with AI

Before adding agent skills or workflow tools, ask AI to assess uncertainty, dependencies, change risk, collaboration, handoff, and existing practices. Starting without extra skills is a valid option. Compare only relevant alternatives, such as Spec Kit, Superpowers, or OpenSpec. Any selected installation and invocation happen in your external coding agent under existing authorization; a chat-only tool can provide manual next steps. Review actual setup results, not a printed command alone.

For the book swap, AI might ask whether a loan needs the owner’s approval and what happens when a book is not returned. Answer unresolved questions. Compare options by your requirements, including those outside this site. Catalog examples are optional. Ask for a recommendation, delegate a specific choice, or explicitly defer it.

A question such as ‘Should loan status be shared across devices?’ connects behavior to data storage and architecture. If the owner marks a book as lent on a phone, should a neighbor see it as unavailable on a laptop? Shared status calls for comparing shared storage and synchronization; a private record used on one device may fit local storage. Ask your external AI to explain these tradeoffs and what should happen offline. This gives you a concrete requirement to compare with the catalog examples, rather than choosing a technology by name.

Ask AI to summarize the users, main actions, failure cases, constraints, decisions, and completion criteria. Check that the plan still describes lending books, and that unresolved questions are visible. Confirm the plan, then explicitly request development. Copying a prompt alone does not approve a plan or start development.

Project planning →

What: a prompt starts the conversation; a catalog supports decisions

kickoff provides learning resources and prompts. The first prompt carries your context and asks your external AI to explore detailed requirements and compare architecture choices; it is not a finished specification. AI conversations and development happen in that external tool. The catalog is optional material for learning and comparing concepts; Project planning provides practical walkthroughs to follow through examples. Your decision list comes from the service, not the category tree. Compatible choices can coexist.

Understand the catalog

Browse this optional index when an example helps. Each link opens a comparison group or subcategories, with examples, limits, and related reading. Neither you nor AI needs to review every category to plan a project.

Explore the catalog →