Cover image for “Agent Skills Hub: finding and managing good Skills”
· 5 min read

Agent Skills Hub: finding and managing good Skills

Also available in 中文

Skills Open source Agent

Agent Skills Hub is a discovery and management platform for Skills. It shipped in under three days, built entirely with vibe coding (AI-assisted development).

Site: agentskillshub.top Open source: github.com/zhuyansen/agent-skills-hub Author: Jason Zhu (@GoSailGlobal)

A hall of skill constellations

Three problems it is trying to solve

1. How do you find the right Skills?

  • there are a lot of them
  • quality is uneven
  • hard to filter

2. How do you find the people who make good Skills?

  • you do not know who is shipping the good ones
  • there is no place to talk

3. How do you combine Skills?

  • a single Skill is limited
  • combinations are more powerful
  • nobody recommends scenes

What it does

  • the Skills people actually use
  • live updates
  • community-checked

2. Skills Masters

  • creators of good Skills
  • portfolios
  • follow and talk

3. Organization Builders

  • teams and orgs
  • series of Skills
  • collaborative development

4. Recently Updated

  • newly published Skills
  • projects that are still maintained
  • a way to follow updates

5. Top Rated

  • highest-quality Skills
  • composite scores
  • user reviews

6. Browse by Category

  • by function
  • by scene
  • find it faster

7. Scenario Workflows

  • Skills used together
  • real scenes
  • practices that work

How it is built

The pipeline

1. Collect

  • pull Skills metadata from many sources
  • GitHub, communities, forums

2. Clean

  • dedupe
  • format
  • check that they still work

3. Score quality

Signals:

  • papers
  • GitHub metrics (stars, forks, issues)
  • token estimates
  • composability
  • a weighted composite

4. Show

  • a front-end
  • search and filters
  • detail pages

Built with vibe coding

Traits

  • entirely vibe-coded
  • front and back generated by AI
  • under three days

Tools

  • Claude Code
  • requirements in natural language
  • the model implements

From the author

“Feel free to roast it. File an issue if something is broken — I’ll have Claude Code do the work 👀“

How to take part

Ways to contribute

1. Submit a good Skill

  • share yours
  • help other people find it

2. Recommend a Skills Master

  • point at good creators
  • grow the network

3. Share a combination scene

  • what you actually ran
  • what worked
  • unusual pairings

4. Star the repo

  • support the open-source project
  • help more people find it

Why it is useful

For users

1. Find good Skills

  • less search time
  • some quality bar
  • clear categories

2. Learn what works

  • see how other people use them
  • recommended combinations
  • ideas you would not have had

3. Find the right people

  • follow Skills Masters
  • learn and talk
  • chances to collaborate

For creators

1. Show the work

  • more exposure
  • feedback
  • a place to build influence

2. Find users

  • see demand
  • collect feedback
  • keep improving

3. Talk to peers

  • meet people in the same craft
  • share experience
  • build together

Combining skill modules

How I would use it

If you are new

First: browse trending

  • see what people actually run
  • learn the mainstream Skills

Second: explore by category

  • start from your need
  • find the matching category
  • try one

Third: learn combinations

  • read the scene recommendations
  • learn how Skills stack
  • get more than one Skill’s worth

If you are further along

First: follow Masters

  • find the good creators
  • watch what they ship
  • steal the thinking, not only the files

Second: contribute

  • share your Skills
  • recommend combinations
  • help the community grow

Third: help build the platform

  • file issues
  • send code
  • improve the hub

What I take from it

On the Skills ecosystem

1. Quality beats quantity

  • more Skills is not better
  • good Skills are the scarce thing
  • you need filtering and scoring

2. Combinations are more powerful

  • a single Skill is limited
  • stacked Skills react
  • scenes are the key

3. The community has to drive it

  • open collaboration
  • build it together
  • keep evolving

On vibe coding

1. Throughput is extreme

  • front and back in three days
  • AI-assisted
  • iterate fast

2. The bar to ship drops

  • you do not need deep programming chops
  • stay on product
  • let AI implement

3. The trend

  • more products will be built this way
  • ideas and demand matter more
  • implementation goes to the model

Comments