Emil Kowalski Releases Design Engineer Skills for AI Coding Tools

Emil Kowalski released three new skills for coding agents. The skills teach tools such as Codex, Claude Code, and Cursor how to apply senior design engineer standards to UI animations.

The announcement came from a single post on X. It described years of collected principles now packaged as reusable agent instructions.

Skills Target Animation Taste Gaps

The skills address a known weakness. Current coding agents produce code that moves elements without clear purpose. They ignore performance costs and accessibility settings.

Kowalski distilled his own practice into three focused skills. Each skill supplies rules that agents can apply during code review or generation.

One skill supplies strict review criteria. Another converts vague requests into precise animation terms. The third enforces the core motion rules during implementation.

Core Rules Agents Must Follow Now

The rules limit when animation appears. Agents must disable motion for any action repeated more than one hundred times daily. They must keep every UI animation under three hundred milliseconds.

Only transform and opacity properties receive animation. Entries start from a scale of zero point nine five combined with zero opacity. Agents must respect the prefers reduced motion setting by removing only the movement portion.

These constraints come directly from production experience. They prevent animation from becoming noise or performance drag.

Review Process Produces Before and After Tables

One skill runs a dedicated animation review. The agent outputs a table that shows the original code, a revised version, and the reason for each change.

The table format forces explicit comparison. It prevents vague feedback such as "make it smoother." Every suggestion must cite one of the published rules.

Developers can run this review on existing codebases or on fresh generations. The output serves as a record for team standards.

Vocabulary Skill Replaces Loose Descriptions

The second skill handles language. Vague prompts like "add a bounce effect" get translated into exact timing, easing, and property choices.

Agents now receive a shared vocabulary that matches professional motion design practice. This step reduces iteration loops when the designer and agent disagree on what "pop" should look like.

The skill also records the new terms so future prompts become shorter and more precise.

Three Skills Used Together in Practice

Teams report applying the full set during feature work. An agent first receives the animation rules, then converts the designer brief into technical terms, and finally reviews the produced code against the same rules.

The sequence creates a closed loop that keeps output consistent even when multiple people or agents touch the file.

Early tests show fewer animation related pull request comments once the skills stay active in the agent context.

Pressure on Existing Agent Workflows

Other coding tools now face direct comparison. Agents that freely apply spring animations or long durations look inconsistent next to the Kowalski rules.

Teams must decide whether to adopt the three skills or maintain their own style guide. The public release lowers the cost of adopting a tested standard.

No competing agent provider has released a comparable packaged skill set at the same level of detail.

Limits and Remaining Questions

The rules focus only on UI interface animation. They do not cover 3D motion, data visualization transitions, or marketing creatives. Teams working outside web interfaces must still define separate standards.

The skills also assume the agent can inspect code and suggest edits. Standalone generation without review still requires human oversight.

Long term performance data across large codebases remains limited. Early adopters will provide the clearest measure of whether the rules reduce motion related bugs over months of use.

What to Watch in the Next Three Months

Track new releases from Cursor and Claude Code that mention built in animation review features. Watch for public templates that bundle the Kowalski skills with project specific additions.

Monitor GitHub repositories that adopt the three skills and publish before and after metrics on animation count or bundle size. Follow developer discussions on X for reported friction or unexpected benefits after wider use.

These signals will show whether the package stays a standalone skill set or becomes a default expectation for agent generated interfaces.

Developers who write or review UI code can test the skills immediately by loading them into any supported agent. The rules are explicit enough to apply even without the packaged version.