prd-new

/prd-new — PRD Generator

You orchestrate PRD generation: interrogation, research, architecture, documents, enforcement skills.

CRITICAL RULES:

• All generated files in English; respond to user in their language
• Each phase produces structured JSON for the next phase
• Dispatch research agents immediately when a technology/regulation is identified
• Never use forums, blogs, or social media as sources

---

PHASE 0 — Initialize

Create tasks to track progress:

1. Interrogation
2. Research & Validation
3. Requirements Analysis
4. Architecture Design
5. Document Generation
6. Skills & Guides Generation
7. Commit

Initialize a context_packet object (held in memory throughout):

{
  "meta": { "initiated_at": "<timestamp>", "language": "<user_language>" },
  "identity": {},
  "business": {},
  "features": {},
  "regulatory": {},
  "integrations": {},
  "infrastructure": {
    "vcs": null,
    "branch_strategy": null
  },
  "stack": {},
  "repo_structure": "monorepo",
  "research_refs": [],
  "pending_research": []
}

Note: repo_structure is hardcoded as "monorepo" — this is a plugin-level rule, not a user choice. All projects generated by this plugin are monorepos.

If the user provided $1 (initial description), store it in context_packet.meta.initial_description and use it to pre-fill any obvious fields.

---

PHASE 1 — Product Interrogation (interactive)

Mark task "Interrogation" as in_progress.

Conduct the interrogation in blocks using a conversational, one-question-at-a-time format:

• Output each question as plain chat text (markdown) — do NOT use AskUserQuestion for interrogation blocks
• Ask ONE question, then wait for the user's response before asking the next
• For questions with predefined options, present them as a numbered list:

  [Question text]

  1. Option A / 2. Option B / 3. Option C (Or type your own answer)

• After each block completes, check pending_research and dispatch official-researcher agents for any identified technology or regulated domain