// starters

sourdough starters.

free catalogue of living context documents for AI coding agents. fork into your own projects, feed as you work, develop character over time. 10 starters so far.

new to the term? read the canonical article → it explains why generic CLAUDE.md templates aren't starters, what the feed ritual is, and why this matters for products built to last.

MIT-licensed. Copy, fork, adapt, ship. No signup needed. Found a shape that works? Contribute back via github — markdown PRs only, low review bar (accuracy + brand-voice fit).

// project

project starter
The canonical session-context document. Five sections every project's starter needs, with the prompts that produce the entries.
last fed 28 may 2026· foundation· canonical· project

// claude.md

claude.md — next.js + supabase
CLAUDE.md template for the most common vibe-coder stack — Next.js App Router + Supabase + Vercel. Stack-specific gotchas, conventions, the "don't do this" list.
last fed 28 may 2026· nextjs· supabase· vercel

// security

security posture
Security defaults that should travel with any vibe-coded project. Secret handling, RLS philosophy, auth check placement, webhook signatures, validation parity — the seven things that account for most of the failures we see.
last fed 28 may 2026· security· rls· auth

// auth

auth patterns
The six recurring auth shapes (where the check lives, library choice + drift, session storage, webhook signatures, RBAC vs ABAC, OAuth state) and the decisions to make once so the AI stops re-deciding them every session.
last fed 28 may 2026· auth· security· session

// data

data modelling
Naming, types, money, identifiers, soft-delete, denormalisation — the schema-shaping decisions that get expensive to reverse. The defaults the AI should treat as constraints.
last fed 28 may 2026· data· schema· postgres

// architecture

architecture decision records (adr)
The format for writing down decisions that need to outlast the conversation. Short, structured, numbered. Catches the "why did we do it this way?" gap that hits every project six months in.
last fed 28 may 2026· architecture· decisions· documentation

// testing

testing strategy
What gets tested, what doesn't, the test pyramid that suits a vibe-coded codebase, and the rules that stop tests becoming the work itself. Optimised for indie / solo / small-team velocity, not enterprise certainty.
last fed 28 may 2026· testing· unit· integration

// ux

ux audit playbook
The eight things to check on every page before it ships. Audit-shaped, not build-shaped — what to look for, not what to add. Catches the issues the AI's preview can't show you.
last fed 28 may 2026· ux· audit· playbook
ux reference
Design tokens, spacing, typography, colour conventions — the visual reference your AI consults before making any UI choice. Replaces the "AI invents arbitrary hex codes" failure mode.
last fed 28 may 2026· ux· design-system· tokens

// brand

brand voice
How your project writes. Tone, conventions, the words you use and don't, the AI tells to strip out. Forks into any project's CLAUDE.md so the AI stops generating prose that reads as AI-generated.
last fed 28 may 2026· brand· voice· copy