Skip to content

Guide · Agent skill

How to use the Starting a New Project agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The starting-a-new-project skill is an agent skill for an empty workspace: a folder with no code and only a project idea. It makes the agent interview you in short rounds, propose a draft for each part, and save the confirmed result as one markdown file, goal-and-requirements.md. This guide explains what the skill does, how to install it, how to trigger it, and what to expect before you begin.

Starting a New ProjectJetBrains/thinkrail · Apache-2.0

Quick start

  1. 1Create an empty folder. Start in an empty workspace with no code, which is the only situation this skill is meant for.
  2. 2Install the skill. Unzip the downloaded ZIP into ~/.claude/skills/ or the project's .claude/skills/, or upload it under Settings → Capabilities → Skills in the Claude apps.
  3. 3Describe your idea. Ask for help scoping the project in plain words, for example that you want to build a small tool for yourself.
  4. 4Confirm the overview. Review the draft overview the agent proposes and correct anything that is wrong.
  5. 5Answer the problem question. Answer the tailored question in your own words and confirm the problem statement the agent writes.
  6. 6Review and approve the draft. Read the full markdown draft of goal-and-requirements.md and approve it so the status changes to active.

What the skill makes the agent do

When you bring a raw idea to an empty workspace, the skill has the agent work from your request rather than asking everything from scratch. It infers a draft overview, asks you a focused question about the problem, and then proposes the smallest useful first version. You confirm or correct each part before it is written to the file.

The skill also tells the agent to cut ideas that are not essential and to offer a search for existing open-source alternatives. The output is a written scope, and the skill deliberately stops there without starting to code.

  • Overview: a short description of what you are building, sized to the idea.
  • Problem: who has the problem, what they do today, and where it breaks down.
  • Capabilities: the features the first version needs, with anything non-essential cut.
  • Optional alternatives: a list of closest open-source projects or products, with a one-line gap and a URL.

Install the skill from the ZIP

Download the ZIP from the skill's page on this directory and unzip it. The package contains SKILL.md. Place the unzipped folder in one of these locations, depending on who should use it:

  • Personal, for all your projects: unzip into ~/.claude/skills/
  • Single project only: unzip into the project's .claude/skills/ folder
  • Claude apps: upload the skill under Settings → Capabilities → Skills

Make sure your agent has what the skill needs

The skill is written for a SKILL.md-compatible agent and refers to specific workflow tools: spec_create, spec_update, edit, ask_user_question, web_search, and fetch_content. It also instructs the agent to read two companion skills, writing-specs and asking-user-questions, before it starts. If your agent lacks these tools or companion skills, the workflow described in SKILL.md will not run as written.

Trigger the skill with a natural request

The skill's description says it applies when the workspace is empty, has no code, and the user brings a raw project idea. Describe your idea in plain words in an empty folder. You do not need a special command. The skill is normally reached through setting-up-a-project, so the agent may route you there first.

  • I want to build a small habit tracker for myself. Let's scope it before coding.
  • We're starting a new public app for freelance invoicing. Help me write the product requirements.
  • This folder is empty. I have an idea for a tool that summarises my meeting notes. Help me define it.

Work through the interview

The agent opens by saying it will nail the goal and scope, then save the result as goal-and-requirements.md. It proposes a draft overview and asks for corrections. Then it asks one tailored question about the problem. Answer in your own words; the agent turns your answer into a statement and asks you to confirm.

Sections are saved as they are confirmed, not in one batch. If you already wrote a detailed brief, the agent can treat the sections you wrote as confirmed and only ask about what is missing. Finally, it shows the full draft in plain markdown for your review. Once you approve it, the agent changes the spec status from draft to active.

  • Personal ideas get a shorter spec with Overview, Problem, Capabilities and Tech Notes.
  • Public or multi-user products get a fuller PRD with target users, goals, non-goals, success conditions and capabilities.

Safety notes: scripts and network access

The listing states that this skill runs no scripts. It does use the network: the agent may run web_search and fetch_content to find existing open-source projects or products for the alternatives section. Expect those lookups if you accept the alternatives offer. The skill does not say what content it fetches beyond that, so review any alternatives list before relying on it.

  • Runs scripts: no
  • Uses network: yes, for alternatives research via web_search and fetch_content

What people use it for

  • Scoping a personal tool, such as a small habit tracker, into a short spec before writing code.
  • Drafting a product requirements document for a public app with named user roles and measurable goals.
  • Deciding what to leave out of a first version by letting the agent cut non-essential ideas.
  • Checking for existing open-source alternatives before you build something from scratch.

Frequently asked questions

Is the starting-a-new-project skill free?

The source lists the licence as Apache-2.0, and that is the only licence information provided. The source does not mention prices, so check the repository's licence file for the terms that apply.

How do I install the starting-a-new-project skill?

Download the ZIP from the skill's page and unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder for one project. In the Claude apps, you can instead upload it under Settings → Capabilities → Skills.

What do I need to run this skill?

You need a SKILL.md-compatible agent with the tools spec_create, spec_update, edit, ask_user_question, web_search and fetch_content. The companion skills writing-specs and asking-user-questions must also be available, since the skill tells the agent to read them first.

Does this skill write code?

No. The skill deliberately produces a single document, goal-and-requirements.md, and ends there. It is not meant for users who want coding to start immediately.

Can I use this skill on an existing project?

No. The skill is only for empty workspaces with no code. For changes to an existing project, the skill says to go through the brainstorming skill instead.

Why isn't the agent starting to build my app?

This workflow is designed to end with the saved spec. After saving, the agent suggests the next step, such as sketching architecture.md and building features through brainstorming, and there is no board or ticket hand-off.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you are starting an app or tool from an empty folder and want a short, confirmed written scope before coding, whether for a personal tool or a public product requirements document. Pick something else if your project already has code, or if you want the agent to start building immediately.

Download the skill

Find exactly what
you're looking for