L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

redneck-engineer

redneck-engineer

Turns a vague idea into a clear build plan — short PRD, task list, user flows, constraints, and acceptance criteria — so Claude Code knows exactly what to build; invoke via /redneck-engineer when you know what you want but Claude keeps building the wrong thing.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: redneck-engineer description: Turns a vague idea into a clear build plan — short PRD, task list, user flows, constraints, and acceptance criteria — so Claude Code knows exactly what to build; invoke via /redneck-engineer when you know what you want but Claude keeps building the wrong thing.

Redneck Engineer

You know what you want; the problem is getting it out of your head and onto paper before anyone starts welding. This skill drags a fuzzy idea into a tight, buildable spec so the next agent stops guessing and starts shipping the right thing.

When to use

  • You have an idea but Claude keeps building the wrong interpretation of it.
  • You want a plan/spec before any code gets written.

Behavior

  1. Interrogate the idea first: ask the 3-5 sharpest clarifying questions that would change the build, and wait for answers before planning.
  2. Write a short PRD — problem, goal, target user, in-scope and explicitly out-of-scope.
  3. Lay out the primary user flows step by step, including the unhappy paths.
  4. List hard constraints: stack, performance, data, deadlines, and anything that must not change.
  5. Break the build into an ordered task list small enough that each item is unambiguous.
  6. Define acceptance criteria — concrete, checkable conditions that mean "done" — for the whole thing and key tasks.
  7. Call out open questions and assumptions so nothing silently gets decided wrong.

Output

A compact, buildable spec — PRD, user flows, constraints, ordered task list, and acceptance criteria — ready to hand straight to an implementing agent.

Skills enregistrées