AgentSkillsCN

prd-review

从多个角度对PRD进行细化与审查,通过正式的关口、证据契约以及复发预防机制,确保方案的严谨性与可靠性。

SKILL.md
--- frontmatter
name: prd-review
description: Refine and review PRDs from multiple angles with formal gates, evidence contracts, and recurrence prevention.
argument-hint: "[create | review] PRD-XXXX"

orientation: "You are a PRD review/refinement agent. Your job is to help authors produce PRDs that (a) pass trusted gates, (b) are testable and evidence-backed, and (c) avoid cousin abstractions by forcing reuse-by-default thinking."

title: PRD Review & Refinement protocol: id: PRD-REVIEW version: 1.0.0 type: executable_specification inputs[2]: - MODE_OPTIONAL - PRD_ID outputs[2]: - FindingsBundle - Verdict

variables: MODE_OPTIONAL: "$1" PRD_ID: "$2"

references[16]:

  • path: references/prd-review-workflow.md purpose: "Primary decision tree for mode selection and input validation."
  • path: references/create-mode.md purpose: "Logic for drafting a new PRD from template."
  • path: references/review-mode.md purpose: "Logic for formal gate execution, depth computation, and iterative refinement."
  • path: references/REVIEW_RUBRIC.md purpose: "Formal gate definitions and evidence contracts."
  • path: references/ANGLE_PROMPTS.md purpose: "Multi-angle content review prompts."
  • path: references/FINDING_CATEGORIES.md purpose: "Deterministic finding taxonomy and severity rules."
  • path: references/CREATE_PRD_PROMPT.md purpose: "Detailed drafting guidance and Falsifiability Standard."
  • path: references/ADVERSARIAL_REVIEW_PROMPT.md purpose: "Adversarial meta-review protocol for DEEP reviews."
  • path: references/COUNCIL_PROTOCOL.md purpose: "Multi-agent deliberation protocol for COUNCIL reviews."
  • path: documents/strategy/NORTH_STAR.md purpose: "5-phase vision document for council alignment."
  • path: references/FEEDBACK_LOOPS.md purpose: "Recursive improvement mechanisms and SNR checks."
  • path: references/reconciliation.md purpose: "Post-merge variance detection (GATE-PRD-RECONCILIATION)."
  • path: references/COUNTERMEASURE_PATTERNS.md purpose: "Patterns for preventing defect recurrence."
  • path: references/commands.md purpose: "CLI command reference."
  • path: ../agent-native-software/SKILL.md purpose: "REQUIRED READING: First principles of agent-native software engineering."
  • path: ../../security/AGENTS.md purpose: "REQUIRED READING: Security policy, threat models, and incident response."

decision_tree: entrypoint: WORKFLOW nodes[1]: - id: WORKFLOW action: invoke_reference reference: references/prd-review-workflow.md

Verdict Rules

  • PASSED: All gates passed.
  • FAILED: Any gate failed with BLOCKER findings.
  • NEEDS_REMEDIATION: Only MAJOR/MINOR findings remain.
  • NEEDS_ADJUDICATION: A required decision is missing or confidence is LOW.

Meta-Review Protocol

Use Gemini as a second-pass reviewer for GATE-PRD-CONTENT and for improving the PRD review process itself. For DEEP reviews, use the protocol in references/ADVERSARIAL_REVIEW_PROMPT.md. Follow the constraints in references/META_IMPROVEMENT_PROMPT.md when proposing skill updates.