Agent Skills, reviewed: independent verdicts before install
We read every file of a skill and weigh six months of public evidence. Each skill gets a score, a verdict and a safety grade. Publishers cannot pay to change them.
Which Agent Skills are worth installing?
The highest verdict scores so far. Each card shows the safety grade and how much evidence sits behind it.
Designer Skills Collection
Best forResearch plans, specs, handoff packages and critiques built on named methods
Poor fitGenerating visually distinctive UI code; Cursor users
Remotion Best Practices
Best forRemotion users who want correct compositions, Studio preview and renders only on request
Poor fitPrompt-to-video without React or a local toolchain
Animate
Best forFront-end developers adding a dropdown, modal, toast, drawer or press effect with fixed curves and durations
Poor fitReact Native motion (routed to animate-expo); critiquing existing motion
Emil Design Engineering
Best forWeb developers using CSS or Motion who want UI animation reviewed against concrete easing, duration and transform rules
Poor fitReact Native or native mobile motion (declined in PR #15)
Frontend Design
Best forNew landing pages, marketing sites or prototypes from a loose brief
Poor fitWork inside an established design system or component library
Brainstorming
Best forClaude Code or Codex developers who want intent clarified and sign-off before code
Poor fitQuick edits and config tweaks where an approval round-trip is friction
Find skills for your work
Ranked lists by job and by agent. Each ranks only skills that fit.
Design and frontend
UI, UX and frontend skills ranked by verdict
Writing
Writing skills ranked by verdict
Claude Code
Skills with a verified Claude Code run
Claude app
Claude skills ranked by verdict
Which Agent Skills work in Claude, Claude Code and Codex?
Skills with a cited run inside each agent. Other support is the publisher's claim.
How does Verify AI Skills review an Agent Skill?
Every verdict follows the same published rules and is tied to one pinned version of the skill.
Read every file
SKILL.md plus every script, hook and reference, at a pinned commit.
Gather evidence
Issues, pull requests, independent evals and discussions from six months.
Score the job
Seven universal measures plus measures for the skill's own job. The worst failure weighs most.
Publish the verdict
A label, a safety grade and a confidence level, with the evidence linked.
Are Agent Skills useful in practice?
How many reviewed skills sit in each verdict band.
Are Agent Skills safe to install?
Our safety grade rates what a skill's files can do.
Which skills should you avoid or use with care?
The lowest verdicts, and why.
What are Agent Skills?
Agent Skills are folders with a SKILL.md file of instructions, plus optional scripts and references. An agent loads a skill only when a task matches its description. Anthropic released the format in October 2025.
my-skill/ ├── SKILL.md name, description, instructions ├── scripts/ optional └── references/ optional
New to Agent Skills?
Short guides on how skills work, how to vet them and how to build your own.
What are Agent Skills?
The SKILL.md format, how agents load it, who supports it.
Are Claude skills safe?
The risks, a vetting checklist and our A to F grade.
Skills vs MCP servers
What each one does and when to use which.
Create your own skill
Build and test a SKILL.md step by step.
Skill not activating?
Why a skill does not fire, and the fixes.
Set up skills in Claude Code
Where skills live and how to manage them.
Common questions
Can a publisher pay for a better verdict?
No. Payment cannot change a score, verdict, safety grade, confidence label or hub position. The rules sit in our editorial independence policy.
How current are the verdicts?
Every verdict page shows its review date and commit, and all current verdicts date from 2026-09-30.
How do I report an error?
Send the page URL, the disputed wording and a primary source. Accepted fixes appear with dates in the public corrections log.