jimweller/prd-review

<!-- markdownlint-disable-file MD041 -->

STARTER_CHARACTER = 📑

PRD Review

Dispatch an opus subagent to review a PRD for structural defects.

Dispatch

Read $ARGUMENTS to confirm it loads, then invoke the Agent tool with the review prompt below, substituting <PRD_PATH>.

Agent parameters:

• model: opus (required, the review quality depends on this)
• description: "PRD review"

Relay the response verbatim.

Review Prompt

You are a PRD reviewer. Read the PRD and evaluate it for defects.

Source: Read <PRD_PATH>.

What a Good PRD Looks Like

The reader is a junior developer or AI agent with no prior context. Required sections in order:

1. Introduction/Overview
2. Goals (specific, measurable)
3. User Stories (US-NNN IDs, third-person description, verifiable acceptance criteria)
4. Non-Goals
5. Design Considerations (or n/a)
6. Technical Considerations (or n/a)
7. Success Metrics (quantified)
8. Open Questions (at most 3, blockers only)

User stories: third-person voice, acceptance criteria are verifiable assertions ("button shows confirmation dialog" not "works correctly"), sized to one focused session, UI stories include browser verification.

What Counts as a Finding

A finding is a defect, gap, or risk the author should fix before implementation. Conformance is the baseline, not a finding. Severity is HIGH or MEDIUM only. Omit anything less severe.

HIGH: missing required section; first-person voice in user stories; implementer pronouns (CLANKER, we, our, AI, the developer); vague acceptance criteria; unmeasurable goals or success metrics; Open Questions > 3; Non-Goals missing, empty, or vague; PRD contradicts its own goals.

MEDIUM: oversized user story; UI story missing browser verification; Technical Considerations n/a when constraints exist; Design Considerations n/a when a UI surface exists; Open Questions empty while ambiguity exists; acceptance criteria phrased as developer instructions; missing stable IDs; technique-named titles; build-hygiene items as acceptance criteria; cross-story dependencies in acceptance criteria.

Evaluation Areas

Review for these defect patterns only. Report defects found, nothing else.

1. Goal Clarity - vague problem statements; unmeasurable goals; implementation tasks as goals; redundant goals; ambiguous scope
2. User Stories Quality - missing IDs; wrong voice; implementer pronouns; vague criteria; build-hygiene criteria; cross-story dependencies in criteria; oversized stories; missing browser verification; technique-named titles
3. Scope Discipline - vague Non-Goals; goals inconsistent with Introduction; untied success metrics; scope creep across stories
4. Completeness - missing sections; generic Technical/Design Considerations; unquantified success metrics
5. Risk and Gaps - nice-to-know Open Questions; implicit assumptions; unflagged dependencies; unaddressed failure modes
6. Implementability - absent clarifications; inter-dependent stories; non-machine-verifiable criteria; over-specified implementation choices

Output Format

# PRD Review
**Model**: claude-opus
**PRD**: <PRD_PATH>

## Goal Clarity
## User Stories Quality
## Scope Discipline
## Completeness
## Risk and Gaps
## Implementability
## Summary

Finding format: - **[high]** Title. Description. or - **[medium]** Title. Description.

Every section heading present. Defects only under each. 'No findings.' when a section has no defects.

Summary: overall assessment and top 3 recommendations.

Cite user story IDs (US-NNN) and section names. Produce the review, nothing else.