PRD Consultive

This skill works in any assistant that reads skills in the SKILL.md format: Claude (claude.ai and Claude Code), ChatGPT, Gemini and coding agents. It doesn't depend on any specific tool. When the environment offers more (clickable questions, files, code execution, visuals in the chat), use it; when it doesn't, the text-only path works just as well.

A good PRD is an idea that has already been challenged. Coding agents don't ask when a document is vague: they fill the gap with a plausible guess and build on top of it with full confidence. So this skill flips the usual order: doubts are settled in the chat, up front, and the final document records decisions, not questions.

The path is always the same, only the depth changes:

idea → understanding → board challenges → convergence → gate (decision sheet) → PRD.md → PRD.html (+ PLAN.md)

1. Language

Speak the user's language. The language of the user's first message sets the language of the whole session: the conversation, the personas' lines, the questions, the decision sheet and every document (PRD.md, PRD.html, PLAN.md, idea.md). These instructions are in English only for convenience; never let English leak into a session in another language. Translate every heading, label, signal name and template text. File names stay as they are (PRD.md, PRD.html, PLAN.md, idea.md).

The profile can fix a language; if it does, that wins. If the user switches language mid-session, follow them from that point on.

Country: infer it from the conversation (language, currency, laws or places mentioned). It changes the local references the board uses: data protection law (LGPD, GDPR, CCPA...), payment methods and currency. If the country matters for a decision (legal, payments) and is still unclear, ask once.