Free tool · no install · no AI tokens
AI Agent Project Starter Kit
Generate a complete .claude/ setup — rules file, a subagent, and a slash command — for Claude Code in one go. Plus a one-paste script to scaffold it all. Runs in your browser.
# Your Project
> Agent guide for Your Project — a web application. Read this before making changes.
## Commits
Use Conventional Commits: `<type>(<scope>): <subject>` (≤72 chars), then a body explaining the *why*. Types: `feat`, `fix`, `chore`, `docs`, `refactor`, `test`.
---
_Keep this file thin and current. When a convention changes, update it here so the agent stays aligned._
---
name: code-reviewer
description: Reviews code changes for bugs, security issues, and style. Use after writing or modifying code.
tools: Read, Grep, Glob
model: sonnet
---
You are a skeptical senior staff engineer reviewing a change.
Rank issues by severity. For each, give: what's wrong, why it matters,
and the smallest fix. Check specifically:
- Edge cases and error handling
- Anything that leaks secrets or PII to a client/bundle
- N+1 queries and obvious performance traps
- Anything that contradicts the project's conventions (read AGENTS.md)
Be blunt. If the change is clean, say so in one line — don't invent
nitpicks.
---
description: Review a file or directory for bugs, security, and style.
allowed-tools: Read, Grep, Glob
argument-hint: [path]
---
Review $ARGUMENTS like a skeptical senior staff engineer.
Rank issues by severity. For each: what's wrong, why it matters, the
smallest fix. Check edge cases, error handling, secret/PII leaks, N+1
queries, and anything that contradicts AGENTS.md. Be blunt; if it's
clean, say so in one line.
…or scaffold the whole thing in one paste
#!/usr/bin/env bash
set -euo pipefail
cat > AGENTS.md <<'AGENTKIT_EOF'
# Your Project
> Agent guide for Your Project — a web application. Read this before making changes.
## Commits
Use Conventional Commits: `<type>(<scope>): <subject>` (≤72 chars), then a body explaining the *why*. Types: `feat`, `fix`, `chore`, `docs`, `refactor`, `test`.
---
_Keep this file thin and current. When a convention changes, update it here so the agent stays aligned._
AGENTKIT_EOF
mkdir -p .claude/agents
cat > .claude/agents/code-reviewer.md <<'AGENTKIT_EOF'
---
name: code-reviewer
description: Reviews code changes for bugs, security issues, and style. Use after writing or modifying code.
tools: Read, Grep, Glob
model: sonnet
---
You are a skeptical senior staff engineer reviewing a change.
Rank issues by severity. For each, give: what's wrong, why it matters,
and the smallest fix. Check specifically:
- Edge cases and error handling
- Anything that leaks secrets or PII to a client/bundle
- N+1 queries and obvious performance traps
- Anything that contradicts the project's conventions (read AGENTS.md)
Be blunt. If the change is clean, say so in one line — don't invent
nitpicks.
AGENTKIT_EOF
mkdir -p .claude/commands
cat > .claude/commands/review.md <<'AGENTKIT_EOF'
---
description: Review a file or directory for bugs, security, and style.
allowed-tools: Read, Grep, Glob
argument-hint: [path]
---
Review $ARGUMENTS like a skeptical senior staff engineer.
Rank issues by severity. For each: what's wrong, why it matters, the
smallest fix. Check edge cases, error handling, secret/PII leaks, N+1
queries, and anything that contradicts AGENTS.md. Be blunt; if it's
clean, say so in one line.
AGENTKIT_EOF
echo "✓ Agent kit written."
Run this from your repo root to create every file with the right paths. Generated in your browser — nothing is sent anywhere.
What's in a good Claude Code setup?
A productive AI-agent project is more than one config file. The starter set is three pieces that work together:
- AGENTS.md / CLAUDE.md — the project rules: stack, commands, conventions, what to avoid. The agent reads this before it works.
- A subagent (
.claude/agents/) — a focused helper the main agent can delegate to (code review, tests, repo exploration) without cluttering the main thread. - A slash command (
.claude/commands/) — a reusable prompt you invoke with/name, so your most-typed instructions become one keystroke.
How to use the kit
- Fill in your project name, type, languages, and commands.
- Pick a subagent and a slash command to include.
- Copy each file to its path — or copy the setup script and paste it at your repo root to create all of them at once.
Frequently asked questions
What does the starter kit generate?
Three correctly-formatted files at once: an AGENTS.md (or CLAUDE.md) project rules file, a subagent in .claude/agents/, and a slash command in .claude/commands/. Plus a one-paste setup script that creates the whole structure for you.
Why generate them together instead of one at a time?
A real Claude Code setup is more than one file — the rules file, a focused subagent, and a reusable command work together. Most generators do one file; this builds the coherent starter set in the right paths so you're productive immediately.
Are the formats correct for Claude Code?
Yes. Subagents use the .claude/agents/ markdown + YAML frontmatter format (name + description required); slash commands use the .claude/commands/ format with description / allowed-tools / argument-hint frontmatter. Both follow the current Claude Code conventions.
Does it use AI or cost anything?
No. It's a template assembler that runs entirely in your browser — no AI, no tokens, no login, no cost. Nothing is sent anywhere.
Want just the rules file? Use the AGENTS.md generator. Want to understand the method behind all of this? Read Claude Skills Explained and Spec-Driven AI Explained.