Skip to content

ibelick/ui-skills (MIT)

Julien Thibeaut

Audit method, a11y ordering, metadata checklist, generation guardrails.

Updated View as Markdown

Upstream: ibelick/ui-skills.

carried6 files, 3 distilled, 2 verbatim, 1 verbatim-minus, each hash-locked to a pinned upstream commit
lives inpe-build 3, pe-review 2, pe-design 1
rulingsL-05, L-06, L-08, L-09 involve this work
watchercompared against upstream weekly; when a carried file changes upstream, an issue opens in the kit

Rulings

L-05 · will-change

animationsdev: apply proactively vs jakubkrehel: only on observed first-frame stutter vs ibelick baseline: never outside an active animation.
Ruling: reactive (jakub) — evidence-driven application only. Baseline’s phrasing survives as the generation guardrail.

L-06 · Custom easing curves

emilkowalski: “built-ins are too weak” vs ibelick baseline: “never introduce custom curves.”
Ruling: split by intent. Custom curves from the kit’s named set: yes. Inventing novel curves ad hoc: no.

L-08 · Disabled submit buttons

ibelick: explain why disabled vs jakubkrehel: never disable; validate on submit.
Ruling: never disable. Wins on accessibility evidence — an enabled button that validates is discoverable; a disabled one is a dead end.

L-09 · Accessibility findings in audits

ibelick improve-ui: “discard a11y findings unless requested” vs every other source.
Ruling: a11y findings are never discarded. improve-ui’s falsification pass survives without its discard rule.

Receipts

Every row links to the upstream file at its pinned commit and to the kit’s copy. Verbatim rows are the same bytes; verbatim-minus rows carry a recorded patch beside a pristine copy; distilled rows are authored from the source and say so.

Navigation

Type to search…

↑↓ navigate↵ selectEsc close