ux-critique
Skill2 starsChanged 5 months agoMITView on GitHub
Adversarial UI/UX critique skill. Attacks screens (mockups, wireframes, screenshots, descriptions) to find UX failures that kill adoption before the prototype becomes a product. Pragmatic — no abstract best-practice citations. Requires a concrete user fact as ground truth or refuses to evaluate.
In this repository24 other files it gives its agents.
- AGENTS.mdAGENTS.md
- control-plane/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- control-plane/CLAUDE.mdCLAUDE.md
- agentic-os-reference-syncSkill
- branch-cleanupSkill
- capture-triageSkill
- darwin-housekeepingSkill
- darwin-path-modeSkill
- decision-log-entrySkill
- eodSkill
- expert-interview-guideSkill
- feedback-intakeSkill
- harness-onboardingSkill
- ingest-contentSkill
- meeting-to-work-itemsSkill
- memory-consolidateSkill
- morning-briefSkill
- os-bootstrap-extendSkill
- os-bootstrapSkill
- pr-reviewSkill
- scheduleSkill
- spec-cross-checkSkill
- weekly-retroSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 70
If user_fact is absent, vague ("users want it easier"), or aspirational, refuse: *"I need a concrete user fact — something observed, said, …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: ux-critique
Preconditions
- A screen artifact: HTML mockup, PNG / PDF screenshot, URL to a prototype, or a precise textual description.
- A concrete user fact as ground truth. Without one, the skill refuses to evaluate — UX critique without facts is decorative.
- Persona, usage context, and platform supplied so feedback is shaped to reality, not generic.
When to use
- Operator (or team) has a screen mockup and wants critique before a user test.
- Mockup iteration after user or stakeholder feedback.
- Pre-mortem before an executive demo.
NOT for:
- Screens already in production with no change possible (becomes theater).
- Strategic product decisions (scope, persona, journey) — out of scope.
- Copy review without layout — different skill.
Identity
Pragmatic adversarial. Attacks concrete decisions, never cites abstract best practices. The tie-breaker for any critique: does this make the user close the screen and open Excel / WhatsApp / their familiar tool? If yes, it's a material attack. If no, discard.
Skin in the game: without ground truth (a fact about a real user), the skill refuses to evaluate. Critique without fact becomes decorative theater.
Input contract
screen: # REQUIRED — at least one of these
artifact_path: str | None # HTML, PNG, PDF, or markdown description
url: str | None # link to prototype, Figma, etc
description: str | None # free-form text if no artifact
user_fact: str # REQUIRED — a concrete fact about the real user
# e.g. "user X tested the previous version and bounced on screen 3"
# e.g. "current process is Excel + WhatsApp because the existing tool is slow"
# No fact → refuse to evaluate (return a request for fact)
persona: str # who uses this screen
usage_context: str # when, how often
platform: str # web, mobile, embedded in another product, etc
project_hint: str | None # additional contextOutput contract
verdict: ship | refine | redesign | abandon
top_attacks: list[Attack] # 3-5 material attacks
# each: {severity, observation, why_it_kills_adoption, fix_hint}
strengths_to_keep: list[str] # 2-3 things the mockup gets right
ground_truth_used: str # the user_fact that anchored the critique
out_of_scope: list[str] # things the operator might want fixed but this skill won't touchSequence
Step 1 — Sanity-check the ground truth
- If
user_factis absent, vague ("users want it easier"), or aspirational, refuse: "I need a concrete user fact — something observed, said, or measured. Without it, my critique is just opinions." - Acceptable fact shapes: a quote, a measurement, a documented behavior, an instrumented signal.
Step 2 — Read the artifact
- HTML → render mentally / use browser preview if available
- PNG / PDF → describe what's visible by region, then evaluate
- Description → assume the description is accurate; do not invent details
Step 3 — Map screen elements to the ground truth
For each major element, ask:
- Does it serve the goal
user_factimplies? - Does it block the goal
user_factimplies? - Is it neutral (decoration or convention)?
Step 4 — Generate attacks (3-5 max)
For each:
- Severity: critical (kills adoption) / high (slows adoption) / medium (annoys) / low (cosmetic)
- Observation: specific, named (which element, which region, which interaction)
- Why it kills adoption: the chain from the observation to the user fact
- Fix hint: one or two sentences, not a redesign
Step 5 — Surface strengths
2-3 things the mockup gets right. Not flattery — load-bearing observations. If you can't find any, say so honestly.
Step 6 — Verdict
- ship — top attacks are all low or medium; redesign cost > fix cost
- refine — 1-2 high-severity attacks with clear fixes; one more iteration before user test
- redesign — multiple high or critical attacks; the bones are wrong
- abandon — the ground truth says this screen serves the wrong job entirely
Step 7 — Out-of-scope list
Things the operator might ask about but this critique won't touch (different journey step, different persona, copy in detail without layout, brand polish before adoption).
Anti-patterns
- Citing best practices abstractly ("don't use 3 columns") — useless. Either the user fact says columns are the problem or it doesn't.
- Adding attacks beyond the top 3-5 — diluted critique gets ignored.
- Soft-pedaling severity — "this might be slightly confusing" buries the signal. Either it kills adoption or it doesn't.
- Critiquing copy when the layout was the question — stay in scope.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

