What is the pm-design-critique skill?
The pm-design-critique skill is the design-critique SKILL.md in the PM Skills library. It asks for the user goal, platform and design stage, then returns a six part critique: strengths, ranked issues, a Nielsen heuristic table, a Gestalt check, a Jobs-to-be-Done check and three next steps.
Its own quality checklist begins with a rule against flattery: the What's Working section may only hold genuine, specific observations. Looks clean does not count, while a hierarchy that guides the eye from headline to call to action does.
The skill is named design-critique in its frontmatter and lives in a very large library of professional skills, each one a plain markdown file. This copy sits in the OpenClaw export folder of that repository.
What it asks you for
- What is being reviewed: a screen, flow, component or full product.
- A design description or an attached image. Without an image, the skill still works from your description.
- The user goal, meaning what the person is trying to accomplish.
- The context, such as web, mobile, desktop app or physical product, and the stage from early wireframe to live product.
- An optional primary concern, for example an onboarding flow that feels too long.
What the critique contains
- What's Working, with 3 to 5 honest and specific observations.
- Priority Issues, each rated High, Medium or Low and tied to a named principle.
- A Heuristic Assessment table scoring Nielsen's 10 usability heuristics as Pass, Partial or Fail.
- A Gestalt check covering proximity, similarity, continuity, figure and ground, and closure.
- JTBD Alignment, which asks whether the primary job is obvious and what distracts from it.
- Top 3 Recommended Next Steps for the next design iteration.
Who the pm-design-critique skill is for
Best for
- Product managers and founders who need written feedback on a wireframe, flow or live screen without a design review meeting. For a plain critique prompt, there is critique.
- Designers who want every comment tied to a principle and not to taste. A related review format is kwp-design-critique.
- Early stage checks, since recommendations are scoped to the stated stage.
Not for
- Running a Figma file review for a product manager. The skill points to a separate figma-design-critique-pm skill for that.
- Replacing user research. It critiques what you describe or attach, and does not test with real users.
- Choosing colors or type. The output is about usability and goals, and not about visual style. See uiux-design for broader design guidance.
Use it with UiChemy
Run the critique on your Figma screen or on a screenshot of a WordPress page, then act on the three next steps. UiChemy Figma to WordPress converts the revised design into an editable page, and Composer handles visual tweaks. Click Get Started.
How to install PM Design Critique Skill: Structured UX Feedback
npx pm-claude-skills addThe README says a skill is a markdown file your AI reads, so installing copies text files and uninstalling deletes them.Run the installer
The README's install line is
npx pm-claude-skills add, described as working for any tool, including Cursor, Codex, Windsurf, Cline and Zed.Or use Claude Code
In Claude Code, run
/pluginand search for pm-skills, which the README says is the official directory route. It also shows bundles such as/plugin install pm-decoders@pm-claude-skills.Or copy this skill only
Clone
https://github.com/mohitagw15856/pm-claude-skills.gitand copy theexports/openclaw/design-critiquefolder into the place where your agent loads skills.
Check that it loaded
Ask your agent to critique a screen and describe the user goal, platform and stage. With the skill loaded the reply should follow six numbered sections and finish with three next steps.
Key features
- Six section critiqueA fixed structure runs from What's Working through Priority Issues, a heuristic table, a Gestalt check and a JTBD check to the top three next steps.
- Ranked issuesEach issue is marked High, Medium or Low by its effect on the user goal, with what is happening, why it matters, a framework reference and a recommendation.
- Nielsen's 10 heuristicsA table covers all ten heuristics and scores each as Pass, Partial or Fail, but only for what is visible in the design and not for things out of scope.
- Gestalt principles checkPrompts the agent to comment on proximity, similarity, continuity, figure and ground, and closure, whether they are well applied or violated.
- Jobs-to-be-Done alignmentAsks whether the primary job is obvious, what competes with it, and which emotional job the design serves, such as speed, confidence, control or delight.
- Stage aware adviceInputs include the stage, from early wireframe to live product, and an anti-pattern forbids recommending a full redesign when the work is already high fidelity.
- No manufactured praiseThe skill says to include only genuine strengths, because invented praise undermines the whole critique.
- Principle over preferenceEvery issue must cite a UX principle or user impact. Leading with a visual preference such as disliking a color is listed as an anti-pattern.
- Built in quality checksA checklist confirms that strengths are specific, issues carry a framework reference, recommendations are actionable and priority levels (High, Medium, Low) reflect actual impact on the user's goal.
- Plain trigger phrasesExample prompts include critique this design, give me feedback on this UI, review this Figma screen for usability issues and do a heuristic evaluation.
Frequently asked questions
What is the pm-design-critique skill?
It is the design-critique SKILL.md inside mohitagw15856/pm-claude-skills, licensed MIT. It gives structured, constructive feedback on a design using Jobs-to-be-Done, Gestalt principles and usability heuristics, with prioritized issues and specific recommendations.
How do I install the pm-design-critique skill?
The README gives npx pm-claude-skills add for any tool, and says to search pm-skills after running /plugin in Claude Code. To take only this skill, clone the repository and copy the exports/openclaw/design-critique folder into your agent's skills location.
What does the output look like?
The skill has a template with a header for the user goal, context and critique focus, then six numbered sections. Issues are rated High, Medium or Low, and the heuristic table scores each item as Pass, Partial or Fail.
Do I need to attach an image?
No. The skill asks for a design description or an attached image, and says it will still work if you only describe the design. An image gives the agent something concrete to critique, so attach one when you can.
Does it work with Claude Code, Cursor or Codex?
The README says the library works with Claude, ChatGPT, Gemini, Cursor and Codex. It lists Claude Code through the plugin command and any tool through the npx installer. This particular copy is the OpenClaw export of the skill.
Does it review Figma files directly?
Yes, as a use case. The skill description covers giving feedback on a Figma file or wireframe, and one example prompt is to review a Figma screen for usability issues. A note in the file also points to a separate figma-design-critique-pm skill for a product manager's critique run against a Figma file.
What anti-patterns does it avoid?
It forbids leading with taste, inventing praise, giving every issue the same priority, skipping the Jobs-to-be-Done section for product screens and recommending a full redesign when the design is already high fidelity.


