audit
Audit current renders and stress coverage, repair real defects and stop under the bound refinement policy.
Review actual current design evidence and rank observable problems without implementing changes.
$ npx -y skills add 3x-haust/oh-my-design --skill critique --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/critiqueContext preview
The summary Claude sees to decide when to auto-load this skill.
Review actual current design evidence and rank observable problems without implementing changes.
name: critique description: Review actual current design evidence and rank observable problems without implementing changes.
Review only; do not patch. Inspect current fixed-view renders across the requested surfaces/states and actual native measurements/checks. Missing pixels or numeric evidence remain unassessed. A screenshot does not prove accessible names, keyboard tasks or screen-reader behavior.
Use the mode-bound Eye packet when independence is claimed. Blind review excludes external references, source, authorship, maker rationale and prior scores. A same-session critique can be useful but is not independent final approval.
Group findings by actual cause and rank by user consequence. Cite localized conditions, required criteria and measurement IDs rather than generic taste. Familiar controls, restrained typography or content-led direction are not defects by themselves. Distinguish required defects from optional polish; do not invent a problem to look thorough.
Return the strongest evidence-backed findings and limitations. If implementation is separately authorized, route copy/type/composition/source repairs to their owners and refresh affected evidence. Review alone does not authorize those edits or a completion claim.
AI that designs like a human. Give OMD one request. It researches, compares real material, chooses a direction, builds, renders, and repairs internally.
Audit current renders and stress coverage, repair real defects and stop under the bound refinement policy.
Interpret recurring measured findings and honest trends without turning taste into a skill score.
Dispatch the existing candidate-generation stage from minimal tokens to one authentic direction choice.
Expand the selected sample into the complete requested surface, state and viewport scope.
Implement an already chosen Figma target faithfully while verifying unrepresented behavior, states and responsive gaps.