audit
Audit current renders and stress coverage, repair real defects and stop under the bound refinement policy.
Repair discourse from verified facts and a stable voice contract without changing meaning or inventing authority.
$ npx -y skills add 3x-haust/oh-my-design --skill humanize --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/humanizeContext preview
The summary Claude sees to decide when to auto-load this skill.
Repair discourse from verified facts and a stable voice contract without changing meaning or inventing authority.
name: humanize description: Repair discourse from verified facts and a stable voice contract without changing meaning or inventing authority.
Identify speaker, listener, situation, intended change, genre/register and exact facts/quotes. In OMD use the copy deck's Voice and Truth contracts; missing consequential facts return to Writer. Read theory/voice.md and, for persuasion, theory/web-copy.md only as needed.
Choose local repair when order works; reconstruct from the same verified facts when the discourse itself is broken. Put the answer/action first, restore concrete actor/action/condition relationships, place qualifiers beside claims and remove repetition that adds no decision. Preserve useful labels, accessible repetition, legal limits and natural language register. Do not optimize for detectors, invent warmth/experience, add quirks or infer authorship.
Compare every proposition with the factual ledger. Preserve uncertainty, numbers, names and exact quotations. UI action/status language stays within Result boundary and Storage boundary. Supporting fact: none is not permission for filler. Korean variable input is data, not a sentence stem with guessed particles.
Only oh-my-design:writer edits .omd/copy-deck.md; run copy --check, obtain a fresh exact-byte copy-editor review when selected, then Hand synchronizes source. Changed visible copy invalidates affected direction inputs, proof and renders. Return the rewrite, with a short audit only when requested or required by the deck. Do not redesign or silently change facts.
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.
Review actual current design evidence and rank observable problems without implementing changes.
Expand the selected sample into the complete requested surface, state and viewport scope.