minimalismlisted

# Minimalism — least code that fully works Modelled on [ponytail](https://github.com/DietrichGebert/ponytail), which measures ~54% less code while keeping every safety property. The goal is the *smallest correct delivery*: fewer tokens, less code, fewer moving parts — never fewer guarantees. You still satisfy every acceptance criterion in full; you just refuse the scope, layers, and abstractions the ticket didn't ask for. This is a lens, not a stage. Apply it while you implement (alongside `plan-change` and whatever build skill the ticket needs), and confirm it in `self-review` before submitting. ## The four questions (ask in order, stop at the first that resolves it) 1. **YAGNI — does this need to exist at all?** Question the task before writing it. Delete speculative code paths, config knobs, options, and "future-proofing" nobody asked for. The cheapest line is the one you don't write. If a whole file/class/flag isn't required by an AC, it isn't in scope. 2. **Standard library before custom code.** Reach for what the language already ships (built-in collections, string/date/JSON utilities, iterators) before hand-rolling it. Don't reimplement what `Array`/`Map`/`itertools`/`stdlib` already does correctly. 3. **Native platform features before a new dependency.** Prefer what the runtime, framework, or browser already provides over adding a package. A new dependency is debt (supply chain, CVEs, build weight) and in this factory needs human approval a