What Is the 'Review-Animations' Skill's Purpose and Bias?

The review-animations skill is a specialized review agent that evaluates animation and motion code against a strict craft standard, biased toward motion that "feels right" and defaulting to flag rather than approve.

The review-animations skill is part of Emil Kowalski's open-source skills repository — a collection of specialized AI agents for design-engineering workflows. This particular skill acts as a high-fidelity gatekeeper for animation quality, ensuring only motion that meets rigorous craft standards passes review.

Purpose of the Review-Animations Skill

The review-animations skill has a narrow, well-defined scope: it reviews animation and motion code exclusively. According to [SKILL.md](https://github.com/emilkowalski/skills/blob/main/skills/review-animations/SKILL.md#L9), the skill does not write new features, fix unrelated bugs, or review non-motion code.

If prompted with a generic code review request, the skill declines and redirects the user toward a general review skill. This boundary enforcement prevents scope creep and maintains the agent's specialized focus.

What the Skill Evaluates

• CSS transitions and keyframe animations
• JavaScript-driven motion (GSAP, Framer Motion, etc.)
• Gesture physics and spring configurations
• Frame budgets and performance characteristics

The "Feels Right" Bias

The review-animations skill carries an explicit bias: toward motion that feels right. This is not a vague preference but a concrete filter with specific criteria.

Animations are expected to be:

• Smooth — no jank, stutter, or frame drops
• Correctly originated — transitions should start from logical trigger points
• Appropriately timed — durations and easing match user expectations
• Frame-budget compliant — performance stays within 60fps constraints

Merely "working" code is insufficient. Sluggish timing, mistimed triggers, or frame drops are treated as regressions, not acceptable trade-offs.

Default Stance: Flag, Don't Approve

The review-animations skill's default posture is to flag potential issues. Approval must be earned through demonstrated craft. This creates asymmetric scrutiny: the burden of proof lies with the code being reviewed.

How to Invoke the Review-Animations Skill

The skill accepts inputs through YAML, CLI-style calls, or JSON payloads. Below are practical examples derived from the repository's workflow conventions.

YAML Invocation

skill: review-animations
inputs:
  diff: |
    @@ -12,7 +12,7 @@
    - transition: opacity 0.3s ease-out;
    + transition: opacity 0.1s linear;

In this example, the skill would likely flag the change: a 100ms linear opacity transition often feels abrupt compared to a 300ms ease-out.

CLI-Style Call

instagit run review-animations \
  --diff "$(git diff HEAD~1 HEAD -- path/to/component.css)"

JSON Payload

{
  "skill": "review-animations",
  "request": {
    "type": "diff",
    "content": "…the git diff shown above…"
  }
}

Each invocation returns a review result that either flags motion quality issues or grants approval when the strict craft bar is satisfied.

Key Source Files

Understanding the review-animations skill requires familiarity with two files in the repository:

File	Role
[skills/review-animations/SKILL.md](https://github.com/emilkowalski/skills/blob/main/skills/review-animations/SKILL.md)	Defines purpose (line 9) and bias statement (line 13)
[skills/review-animations/STANDARDS.md](https://github.com/emilkowalski/skills/blob/main/skills/review-animations/STANDARDS.md)	Optional detailed criteria for review enforcement

The SKILL.md file contains the canonical definition: lines 3, 9, and 13 establish the skill's identity, scope boundary, and evaluative bias respectively.

Summary

• review-animations is a specialized review agent, not a general-purpose tool
• Its purpose is strictly limited to evaluating animation and motion code
• Its bias demands motion that "feels right" — smooth, well-timed, performant
• It defaults to flagging and only approves when craft standards are met
• It declines out-of-scope requests and redirects to appropriate skills

Frequently Asked Questions

What happens if I ask review-animations to review non-animation code?

The skill declines and points you toward a general review skill. Its scope is strictly limited to motion code per [SKILL.md line 9](https://github.com/emilkowalski/skills/blob/main/skills/review-animations/SKILL.md#L9).

Does review-animations ever write or fix code?

No. The skill is review-only. It evaluates existing code against standards but does not generate new features or apply fixes. This separation of concerns keeps the agent's role unambiguous.

How strict is the "feels right" standard?

Very strict. The bias treats functional-but-kludgy animations as regressions. Code that "works" but feels sluggish, mistimed, or drops frames will be flagged. Approval requires demonstrated craft in timing, easing, and performance.

Can I override the default flag stance?

The skill's architecture does not expose a "lenient mode." The asymmetric default — flag unless proven worthy — is intentional. To gain approval, revise the animation to meet the criteria in STANDARDS.md or justify why the flagged behavior is actually correct.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}