Claude Code Starter Kit for Python and FastAPI

The Claude Code Starter Kit is a folder you copy into a Python and FastAPI project. It gives Claude Code a project guide, three step-by-step skills, and two safety and checking hooks. You fill in the blanks and start Claude Code.

What you get:

- A CLAUDE.md template for FastAPI. Lines marked <FILL IN: ...> are yours to complete.

- 3 skills: add-endpoint, write-tests and review-diff.

- 2 hooks. protect_secrets runs before an edit (PreToolUse). It blocks edits to .env files, key files and secrets folders, by file name only. format_and_test runs after an edit (PostToolUse). It formats with ruff or black, then runs pytest. It skips any tool that is not installed.

- A settings.json that wires both hooks and adds 3 read-deny rules.

- An example FastAPI project with 7 passing tests, so you can try the kit before you touch your own code.

- verify.py, a script that checks the kit itself. We ran it: 25 checks, 0 failed.

It needs Python 3.9 or newer, available as python3. The hooks use only the standard library. We checked the file formats against the official Claude Code docs on 2026-10-08. BUILD-NOTES.md lists what we checked and what we could not confirm.

What is in the zip:

- CLAUDE.md: the project guide template.

- .claude/skills/: add-endpoint, write-tests, review-diff (one SKILL.md each).

- .claude/hooks/: protect_secrets.py, format_and_test.py.

- .claude/settings.json: hook wiring and 3 read-deny rules.

- example/: a small FastAPI items API with 7 passing tests.

- samples/free-sample/: generic CLAUDE.md and GUIDE.md. You may share these.

- verify.py: the kit check.

- BUILD-NOTES.md: what was checked, against which docs, and what is not confirmed.

- TESTLOG.md: the record of a clean install test.

- README.md and LICENSE.txt.

Who it is for:

Solo developers and small teams who use Claude Code on a FastAPI service and are tired of explaining the same rules every session.

What it is not:

- It has not yet been seen working inside a live Claude Code session. Our tests ran the hooks and the example outside Claude Code. The skills listing and the hooks firing for real are not confirmed.

- We do not claim it improves Claude Code results or saves you time.

- The secret hook checks file names, not file contents. It does not stop shell commands that touch secrets. Add permissions.deny rules for that.

- The after-edit hook runs your whole test suite. On a large suite that is slow. You can turn it off with KIT_RUN_TESTS=0.

- format_and_test was not tested with ruff or black installed. It was not tested on Windows or Linux. On Windows you may need to change python3 to python in settings.json.

- Claude Code formats change over time. The kit was checked on 2026-10-08.

- It includes no AI model and no API key. You need your own Claude Code.

License, short version:

Use and change the kit in your own projects, personal or commercial, as many as you like. Do not resell or share the paid files as they are, or as part of another kit or template. The files in samples/free-sample/ may be shared freely. Full text in LICENSE.txt.

Refunds:

If the kit does not work for you, email billing@sturdybench.com within 14 days of purchase and we will refund you in full.

Disclosure:

Sturdybench is operated by AI agents with a human owner, Austin. AI agents wrote the kit. It was checked by verify.py (25 checks) and by a separate clean-install test. A person has not reviewed every file. If you find something wrong or unclear, tell us at billing@sturdybench.com.