SuperAPI AI Directory

review-prd

General & other

View repository

Updated 10-15 22:16

About this skill

Reviewing a PRD

Evaluate a PRD for quality — not agreement with a preferred solution.

HARD-GATE

  • Review the document, not the idea.
  • Every finding must cite a specific PRD section (redact sensitive data).
  • No alternative solutions unless a requirement is infeasible.

Core Process

  1. Receive the PRD.
  2. Apply the review checklist (completeness, testability, clarity, feasibility, scope, dependencies, edge cases).
  3. Classify each finding as Critical, Suggestion, or Note.
  4. Produce a findings table with severity, evidence, and recommendation.
  5. Verdict: Approved / Approved with Suggestions / Needs Revision.
  6. If Needs Revision, instruct author to address Critical findings and re-run review.

Output

  • Verdict.
  • Findings table with columns: #, Severity, Section, Finding, Evidence, Recommendation.
  • Summary with count by severity.
  • Acknowledgement of well-written sections.

Review Checklist (abbreviated)

Completeness: goals and success metrics are measurable; out-of-scope explicit. Testability: requirements can be verified with pass/fail tests; acceptance criteria are specific. Clarity: no undefined jargon; no ambiguous modal verbs. Feasibility: timeline realistic; dependencies identified. Scope & Edge Cases: failure scenarios described; migration/compatibility concerns.

Refer to the full checklist in the skill’s reference documentation.

Integration

Skill When
create-prd Review after drafting
plan-tickets Draft tracker-ready tickets after review passes
work-router with tech-lead intent Add a feasibility focus to the review; it does not select a separate persona