improve-animations Skill: Audit App Motion and Get Fix Plans
Makes your agent audit the animations in a codebase and write step-by-step fix plans that any other agent can carry out.
- Installs
- 240.8K
- on skills.sh
- Stars
- 52.7K
- on GitHub
- Files
- 2
- besides SKILL.md
- Updated
- Oct 2026
- last repo commit
The skill acts as a senior motion advisor. It maps where animation lives in your code, audits it against eight categories, shows a table of findings ordered by impact, and then writes one plan per finding you choose. The plans go in a plans/ folder (or animation-plans/ if plans/ is taken).
The audit modes read code and write only those plan files. The one mode that changes source code is execute <plan>, which runs in an isolated worktree. When invoked with no question, it answers with a single line saying it is ready and waits for your request. Its rule values are in the bundled AUDIT.md, based on Emil Kowalski's animation approach.
Good for
- Front-end developers whose UI feels sluggish or off
- Teams that want a ranked list of motion fixes before they spend time
- Anyone who plans with a strong model and codes with a cheaper one
Install improve-animations
Claude Code
- 1
Copy this prompt.
Install the improve-animations skill in Claude Code. Run this command: npx skills add https://github.com/emilkowalski/skills --skill improve-animations Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js.
From its listing on skills.sh. Source.
What's in the skill
When it loads
Survey a codebase's animation and motion code as a senior motion advisor, then produce a prioritized audit and self-contained implementation plans for other agents (or cheaper models) to execute. Read-only on source code — it plans improvements, it does not apply them. Use when the user asks to "improve the animations", "audit the motion", "make this app feel better", or wants a roadmap of animation fixes rather than a review of a single diff.
Sections in SKILL.md
- Initial Response
- Operating Posture
- Hard Rules
- Workflow
- Phase 1 — Recon (always first)
- Phase 2 — Audit (parallel)
- Phase 3 — Vet, prioritize, confirm
- Phase 4 — Write plans
- Invocation Variants
- Tone
Key facts
- Author
- Emil Kowalski
- Installs on skills.sh
- 240,845 (Nov 2, 2026)
- Last repo commit
- Oct 28, 2026
What it includes
Maps stack, motion libraries, tokens, personality and how often each animation is hit
Greps for transition, animation, keyframes and similar. The frequency map drives severity.
Files ·
Phase 1: ReconChecks 8 categories, from purpose and easing to accessibility and missed opportunities
Categories: purpose and frequency, easing and duration, physicality and origin, interruptibility, performance, accessibility, cohesion and tokens, missed opportunities. Larger repos fan out to read-only subagents.
Files ·
Phase 2: AuditRe-checks each finding in the code, shows one ranked table, waits for your pick
Severity is HIGH, MEDIUM or LOW. Non-interactive runs default to the top 3 to 5.
Read ·
Phase 3: Vet and confirmWrites numbered plan files into plans/ plus a plans/README.md
Each plan is stamped with the current commit and holds exact paths, values, steps and a feel-check.
Runs code ·
Phase 4: Write plansSets audit depth: high-traffic components only, or the whole repo
The default is standard. quick gives about 5 high-severity findings.
quick / deepAudits one category, such as performance, accessibility or easing
<category focus>Skips the audit and writes one plan for the improvement you describe
Write ·
plan <description>Has a subagent carry out a plan in an isolated worktree, then reviews the diff
This is the one mode that changes source code, and it happens in a separate worktree.
Runs code ·
execute <plan>Re-checks existing plans against the code and marks done or stale ones
Write ·
reconcile
From the official docs
Permissions
Low risk
The audit only writes plan files, but the execute mode lets a subagent edit your code in a worktree. Use it on a clean branch.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only modeHow to turn it on
Watch out for
The execute mode edits your code
It runs a subagent in an isolated worktree. Start from a clean branch and read the diff.
It creates files in plans/
Add plans/ to your review or .gitignore as you prefer. The skill uses animation-plans/ if plans/ is already in use.
Technical details
| Turning on read-only | Audit modes only write plan files under plans/; avoid the execute mode. |
|---|
Questions
What does improve-animations do?
It audits the animation and motion code in your project and writes prioritized fix plans. Each plan is detailed enough for another agent, including a cheaper model, to follow with no extra context.
How do I install it in Claude Code, the Claude app or Codex?
Use the install box above. SKILL.md does not list specific clients.
Does it change my code?
No, not in the audit. It writes plan files under plans/. Only the execute mode changes code, and it does so through a subagent in an isolated worktree.
Does it need an API key?
The SKILL.md mentions none.
Is it safe?
It bundles no scripts (the folder holds SKILL.md, AUDIT.md and PLAN-TEMPLATE.md). It writes files into plans/, runs git rev-parse --short HEAD to stamp each plan, and in execute mode it dispatches a subagent that edits code in a worktree, so use execute only on a clean branch. SKILL.md also tells the agent to treat repository text as data, not instructions.
How is it different from review-animations?
The SKILL.md says review-animations reviews a single diff. improve-animations surveys the whole codebase and produces a roadmap.
Information
updated Nov 2, 2026
- Repository
- emilkowalski/skills
- Folder
skills/improve-animations
Sources
Docs, issue trackers and discussions this page is built from.
Show all 2Hide
- emilkowalski/skills improve-animations SKILL.md · github.com
- review-animations SKILL.md · github.com