design-review-orchestrator

Routes a non-coding design review through the smallest useful sequence of framing, critique, exploration, specialist perspectives, and direction selection. Use when a broad or uncertain design request

SKILL.md

GitHub

Design Review Orchestrator

Purpose

Turn a broad, vague, or contested design request into a proportionate review process and one defensible design recommendation. Select only the skills and stages that add decision value; do not turn the review into code or an implementation plan.

When to use this skill

Use this skill when:

  • A user asks for a comprehensive design review, redesign process, design workshop, or coordinated use of design skills.
  • An existing page, mockup, or concept feels generic, weak, unclear, or difficult to improve decisively.
  • Several perspectives or alternatives need to be reconciled into one direction.
  • The user is unsure whether the work needs framing, critique, exploration, or a decision.

Do not use this skill when:

  • One focused request is already covered by a specialist skill, such as a layout critique or a decision between developed directions.
  • The user wants frontend implementation, CSS, technical architecture, a dedicated accessibility audit, or copywriting.
  • There is no design question beyond a small local correction.

Inputs to inspect

Inspect the smallest useful set of available inputs:

  • The request, desired outcome, audience, primary message or action, constraints, and stated preferences.
  • Existing screenshots, mockups, wireframes, reference sites, design-system guidance, content, or repository evidence.
  • Existing candidate directions and prior review notes.

State consequential assumptions. Do not invent brand, audience, or business objectives that would change the recommendation.

Workflow

  1. Diagnose the decision. Determine whether the main need is framing, critique, exploration, a specialist lens, comparison, or a combination. Identify the decision the process must enable.
  2. Select the smallest route. Use the patterns below as guides, skipping stages that would only repeat settled work:
    • Unclear brief or purpose: design-problem-framer.
    • Existing composition with a known concern: layout-principles-review.
    • No viable alternatives: layout-direction-explorer.
    • Two or more developed alternatives: design-direction-decision.
  3. Add specialist perspectives only for a material tension. Choose at most the lenses that answer an unresolved question:
    • distinctiveness or a predictable concept: provocative-concept-director;
    • typography, rhythm, or content-heavy composition: editorial-digital-designer;
    • signal reduction, restraint, or identity coherence: minimalist-brand-director;
    • offer clarity, trust, or calls to action: conversion-creative-director;
    • time, feedback, or experiential pacing: motion-experience-director;
    • an ambitious interaction whose feasibility affects the direction: creative-technologist.
  4. Keep perspectives independent. Give each selected specialist the same design frame and evidence. Record its unique insight, tradeoff, and any assumption; do not present separate personas as a final answer.
  5. Challenge and converge. Compare the viable directions against purpose, audience, message clarity, hierarchy, fit, distinctiveness, and constraints. Separate a strong concept from a polished execution. If a tradeoff remains unresolved, name it rather than averaging conflicting advice.
  6. Choose and bound the recommendation. Recommend one primary direction. A hybrid is valid only when it has a clear base and one to three reinforcing elements borrowed from another direction. State what to preserve, change, reject, and investigate next.
  7. Stop before implementation. Return a design decision or a clear next design workflow. Do not write code, create files, modify tokens, or create an implementation backlog unless separately requested.

Orchestration rules

  • Move from understanding to exploration only when the design frame is sufficient to judge alternatives.
  • Critique an existing design before replacing it; preserve elements that serve the objective.
  • Use divergence to create meaningfully different options, not cosmetic variations.
  • Do not call every available specialist or extend the process after a clear winner emerges.
  • When related skills are unavailable, apply the relevant reasoning directly and disclose that no separate skill was invoked.
  • Treat accessibility, responsiveness, and feasibility as constraints when they affect the direction, but route detailed audits or technical planning to their dedicated workflows.

Output format

Return:

# Design Review

## Current state

- **Decision needed:** ...
- **Evidence reviewed:** ...
- **Assumptions:** ...

## Workflow selected

1. <skill or analytical stage> — <why it is needed>
2. ...

## Diagnosis

- **What works:** ...
- **Main problems:** ...
- **Decisive constraints:** ...

## Specialist perspectives

- **<lens>:** <unique insight and tradeoff>

## Directions considered

- **A — <name>:** <strength and risk>
- **B — <name>:** <strength and risk>

## Recommendation

**Choose:** <primary direction>
**Confidence:** High | Medium | Low

<Why this direction best meets the decision criteria.>

## Preserve, change, and avoid

- **Preserve:** ...
- **Change:** ...
- **Avoid:** ...

## Remaining design questions

1. ...

## Next phase

<The next design workflow, or state that the decision is ready for separate implementation planning.>

For a small request, use a compact version containing the diagnosis, selected route, recommendation, preserve/change/avoid guidance, and next phase.

Quality bar

The task is complete only when:

  • The selected workflow is proportional to the uncertainty and impact of the decision.
  • Each included specialist or stage has a distinct purpose.
  • The recommendation is singular, evidence-led, and explicit about tradeoffs and uncertainty.
  • The result distinguishes reviewed evidence from assumptions and does not claim unavailable skills were invoked.
  • The response ends with a design decision or a justified next design step, not an open-ended list of ideas or implementation work.

Edge cases

  • One existing design, no stated goal: frame the problem first; give only provisional critique where the objective is unknown.
  • Several screenshots: compare only those solving the same problem; separate unrelated concepts before deciding.
  • Strong disagreement: expose the strategic choice and weight it against the project objective rather than averaging perspectives.
  • Reference sites: extract transferable principles, not a derivative visual recipe.
  • Accessibility or technical risk: flag the design-level implication and recommend the appropriate dedicated review; do not present it as a completed audit.

Related skills

  • design-problem-framer
  • layout-principles-review
  • layout-direction-explorer
  • design-direction-decision
  • creative-direction-orchestrator
  • provocative-concept-director
  • editorial-digital-designer
  • minimalist-brand-director
  • conversion-creative-director
  • motion-experience-director
  • creative-technologist