Turn rough notes into an engineer-ready PRD

Solo PMs at seed/Series A startups, founder-engineers who need to spec features for contractors, and non-PM founders who don't want to hire a PM yet · Assemble in ~10 min; first draft PRD in <5 min once installed; full reviewed-and-refined PRD in ~20 min.

ProductEdit flowon X● golden 96

step 1connector

Collect and normalize raw input

⚡ bring your own tool

step 2ready

Interrogate the idea (YC office hours)

ent-office-hours

step 3ready

Draft the full PRD

prd-development

step 4ready

Adversarial persona review

meeting8 alts

step 5ready

Refine PRD via parallel analyst team

refine-prd

step 6ready

Risk and resource analysis

analyze-prd

step 7ready

Engineering handoff prompt

prp

step 8ready

Design handoff brief

design-handoff-brief3 alts

Did this recipe work for you?

What you get

▸Refined, production-ready PRD (problem statement, user stories, acceptance criteria, edge cases, success metrics)
▸Atomic ticket manifest engineers can pick up directly
▸Risk assessment and T-shirt-sized resource estimates
▸Engineering handoff prompt (PRP) and design handoff brief ready to paste or share

tools: No paid tools required — all steps run in Claude Code with harvested skills

tokens: ~$0.50–2.00 per PRD session (steps 3–7 are the heavy passes)

≈ $5–20/mo for a team doing 5–15 PRDs per month

What stays yours

· Confirm the clarified problem statement after step 2 — the YC interrogation surfaces reframes that need founder sign-off before drafting begins
· Review and approve the refined PRD after step 5 before handing off to engineers — the AI panel catches gaps but cannot make product tradeoff decisions
· T-shirt size estimates (step 6) are rough signals only; a technical lead should validate before committing to a sprint

✓ Last verified 2026-06-11claude-sonnet-4-6 (subagent, test-mode fixtures)

✓ step 1 · Collect and normalize raw input

✓ step 2 · Interrogate the idea (YC office hours)

✓ step 3 · Draft the full PRD

✓ step 4 · Adversarial persona review

✓ step 5 · Refine PRD via parallel analyst team

✓ step 6 · Risk and resource analysis

✓ step 7 · Engineering handoff prompt

✓ step 8 · Design handoff brief

view sample outputs from this run

step 2 · Interrogate the idea (YC office hours)

# Clarified Problem Statement
*YC Office Hours Interrogation — applied to raw-idea-brief.md*

---

## 1. Who is actually in pain?

Not engineers. **Code reviewers** — the senior engineers and tech leads who receive PRs and must reverse-engineer intent from diffs. They burn 30–90 minutes per PR asking "what does this do?", "why now?", "is this safe to merge?"

The PR *author* feels no acute pain at the moment of writing — they already know the context. The pain lands asymmetrically on the reviewer.

Secondary sufferer: the PM / release manager who tracks what's going into a release and currentl…

step 3 · Draft the full PRD

# PRD: PR Context Engine (PRCE)
*Version 0.1 — Draft*
*Based on: clarified-problem-statement.md*

---

## Overview

| Field | Value |
|---|---|
| Product name | PR Context Engine (PRCE) |
| Status | Draft |
| Author | PM (Ana) |
| Last updated | 2026-06-11 |
| Target release | Q3 2026 |

---

## 1. Problem Statement

When a pull request is opened on GitHub Enterprise, the reviewer — typically a senior engineer or tech lead — has no structured context about the intent, risk, or scope of the change. They must either interrupt the author, read the entire diff cold, or merge blindly.

**Impact (me…

step 4 · Adversarial persona review

# Adversarial Persona Review — PR Context Engine PRD
*Panel: PM, Senior Engineer, Designer, Skeptic User*
*Source: draft-prd.md*

---

## Panel Composition

| Persona | Name | Background |
|---|---|---|
| PM | Jordan | 8 years product, ex-Atlassian, shipped developer tooling at scale |
| Senior Engineer | Priya | Staff engineer, 12 years, has managed mono-repo infra at 500-person org |
| Designer | Marcus | Design systems + dev-tool UX, ex-Linear |
| Skeptic User | Tom | Senior engineer, 14 years, resistant to "AI slop" in dev workflow |

---

## Round 1: Problem Framing

**Jordan (PM):**
The …

The problem

When I have a feature idea sketched out in Slack messages, a meeting recording, or bullet notes, I want an AI to expand it into a complete PRD with problem statement, user stories, acceptance criteria, edge cases, and success metrics, so that engineers can start immediately without another round of clarifying meetings.

Cybernetic layers

Memorybyo

PRD version history and decision log. No harvested memory artifact yet — save each version to a local file or Notion page and reference it by filename in subsequent steps. This is the #1 gap: a PRD-aware context file that tracks open questions and decisions across sessions.

Backgroundgap

All steps are synchronous and human-paced — there is no background/cron layer needed for this job. The refine-prd workflow (step 5) fans out internally; no external scheduler required.

Harnessready

Claude Code is the runtime. Install each ready skill into ~/.claude/skills/ and invoke by name. The refine-prd workflow runs as a .js workflow file in .claude/workflows/. Cowork can run any step in a shared session.

Click any step to see what fills it — and the ranked alternatives you can swap in.