Purpose
Produce a ready-to-run critique framework for a specific design artifact — evaluation criteria, pointed critique prompts, a severity-tagged feedback structure, and a facilitation script — so design reviews stay rigorous, consistent, and focused on craft rather than opinion. Eliminates vague "I don't love it" feedback and gives the reviewer a repeatable structure that holds a high bar.
When to Use
Invoke this skill when the user asks to:
- Prepare a structured critique for an upcoming design review
- Build evaluation criteria for a specific screen, flow, component, or prototype
- Standardize how their team gives design feedback
- Turn a vague review into a rigorous, criteria-based one
- Create critique prompts that push a design toward higher craft and clarity
- Onboard a new designer to how the team evaluates work
Inputs Required
Before starting, collect:
- Artifact type: What's being reviewed — a single screen, a full flow, a component/pattern, a prototype, a design system contribution, or brand/visual exploration
- Design stage: Where it is — early exploration, mid-fidelity, high-fidelity, or pre-handoff
- Product context: What the product does and who uses it
- Design goal: What this artifact is trying to accomplish (the problem it solves, the metric or behavior it targets)
- Constraints: Platform, brand system, accessibility requirements, technical limits
- Review format: Async written critique, live synchronous session, or 1:1
- Bar to hold: Is this a directional check or a ship-quality gate?
If review format or design stage is missing, default to a live synchronous session at mid-fidelity and note the assumption. Criteria weighting shifts by stage — see Notes.
Instructions
Step 1: Frame the Critique
Set the terms of the review before evaluating anything.
- Restate the design goal in one sentence — this becomes the anchor for every critique. Feedback that doesn't ladder up to the goal is noise.
- State the stage-appropriate bar: early exploration is judged on range and concept, not pixel polish; pre-handoff is judged on completeness and craft, not concept.
- Name what is out of scope for this review so the session doesn't drift (e.g., "we are not relitigating the IA today").
Step 2: Build Evaluation Criteria Across 7 Dimensions
For each dimension, produce (a) what "good" looks like at this artifact's stage, and (b) 2–4 targeted critique prompts phrased as questions the reviewer asks the work — not the person.
Dimension 1 — Goal Alignment Does the design solve the stated problem for the stated user?
- Prompts probe whether the primary action is obvious, whether the design serves the user's job-to-be-done, whether success is measurable.
Dimension 2 — Information Hierarchy & Clarity Can the user understand what they're looking at and what to do next within seconds?
- Prompts probe scannability, the single most important element on the screen, competing focal points, and whether the eye path matches the intended priority.
Dimension 3 — Interaction & Flow Does the experience move correctly through states and steps?
- Prompts probe empty/loading/error/success states, edge cases, back-navigation, and whether any step could be removed.
Dimension 4 — Visual Craft & Consistency Is the execution precise and system-consistent?
- Prompts probe spacing rhythm, alignment, type scale adherence, color usage against tokens, and off-system one-offs that create maintenance debt.
Dimension 5 — Accessibility & Inclusivity Is the design usable by everyone in the audience?
- Prompts probe color contrast ratios (WCAG AA: 4.5:1 body text, 3:1 large text), touch target size, keyboard/screen-reader paths, and reliance on color alone to convey meaning.
Dimension 6 — Content & Copy Does the language do its job?
- Prompts probe whether microcopy is clear and specific, whether labels match user vocabulary, whether error messages tell the user how to recover, and whether tone fits the brand.
Dimension 7 — Feasibility & Handoff Readiness Can this be built as intended, and is it ready to hand off?
- Prompts probe responsive behavior, technically expensive choices, undocumented states, and whether an engineer could build this without guessing. (Weight this heavily only at high-fidelity/pre-handoff.)
Step 3: Define the Severity Taxonomy
Give the reviewer a shared vocabulary so feedback is prioritized, not flattened:
- 🔴 Blocker — Prevents the design from meeting its goal or shipping. Must be resolved.
- 🟡 Consider — Meaningful craft or clarity issue worth addressing; not a blocker.
- 🟢 Polish — Minor refinement or nice-to-have.
- 💡 Idea — Expansive suggestion or alternative direction, offered without obligation. Every piece of feedback in the session gets a tag. This prevents a color-preference comment from carrying the same weight as a broken flow.
Step 4: Write the Facilitation Script
Produce a step-by-step run-of-show sized to the review format:
- Set context (2 min) — designer states the goal, stage, and the specific decision they want feedback on.
- Silent review (3–5 min for live) — reviewers examine the work against the 7 dimensions before anyone speaks, to prevent anchoring on the first voice.
- Structured pass — walk the dimensions in order; capture feedback with severity tags.
- Designer synthesis — the designer restates what they heard and names their next moves. They own the decisions; critique informs, it doesn't dictate.
- Action capture — log blockers and owners with a next check-in.
Include 3–4 facilitation guardrails (e.g., "critique the artifact, not the author," "one conversation at a time," "no solutioning during the structured pass — flag the problem, let the designer solve it").
Output Format
- Structure: Framing section → 7 dimension blocks (each with "what good looks like" + tagged critique prompts) → severity taxonomy → facilitation script
- Length: 1,000–1,400 words
- Format: Structured document — dimensions as labeled blocks with bulleted prompts; facilitation script as a numbered run-of-show
- Sections: Critique Frame → 7 Evaluation Dimensions → Severity Taxonomy → Facilitation Script → Guardrails
- Tone: Design manager — exacting on craft, generous to the person; direct, specific, never precious
Quality Checklist
Before delivering output, verify:
- [ ] All 7 dimensions are present, each with a stage-appropriate "what good looks like" and 2–4 critique prompts
- [ ] Every critique prompt is phrased as a question about the work — none critique the designer
- [ ] Criteria are calibrated to the stated design stage (early exploration is not judged on pixel polish)
- [ ] Severity taxonomy has exactly 4 tiers, each applied consistently in the facilitation script
- [ ] Accessibility dimension cites specific, checkable thresholds (contrast ratios, target sizes) — not "make it accessible"
- [ ] Facilitation script is sized to the review format provided and includes a silent-review step to prevent anchoring
Notes
- Criteria weighting shifts by stage: at early exploration, weight Goal Alignment and Interaction & Flow; at pre-handoff, weight Visual Craft, Accessibility, and Feasibility. Applying pre-handoff scrutiny to an early concept kills momentum and range.
- The single most common failure in design review is unweighted feedback — the severity taxonomy is what keeps a spacing nitpick from derailing a session with a broken flow.
- For async written critique, drop the silent-review timer but keep the structured dimension pass and severity tags — they matter more in writing, where tone is easily lost.
- The designer owns the decision. A critique framework's job is to raise the quality of the questions, not to hand down verdicts.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills