Skip to content

Skill Management

A Skill is an installable capability package for AI tools, usually centered on SKILL.md. PromptHub helps collect, preview, translate, install, and maintain these skills.

Skills can also appear in Explore Community. Community skills work like a public capability directory: read what the skill does, then decide whether to install or adapt it.

Capability Description
Skill library Central place for imported and built-in skills
Local scan Discover existing SKILL.md files
One-click install Distribute skills to target AI tool directories
Copy / Symlink Choose stable copies or live linked files
AI translation Read English skills with translated sidecar content
Tags Filter skills by tool and use case
Location Purpose Recommended use
Local skill library Your own capabilities, team rules, and project skills Keep private when they include internal context
Community skills Public skills shared by other users Read, copy, adapt, then install
Official skills Curated entries maintained by the official team Still review tool fit and project fit before use

Do not treat a community skill as your team standard immediately. Copy it into your own library, remove rules that do not apply, and then install it.

Before installing a community skill, check:

  1. What problem it solves.
  2. Which AI tool it targets.
  3. What input it expects from you.
  4. What output it produces.
  5. Which language, framework, platform, or team assumptions it contains.

If the description is abstract and does not explain input, output, or limits, do not install it directly.

Copy mode is stable and self-contained. Symlink mode keeps targets connected to the source file, but depends on filesystem permissions and stable paths. Choose copy when you are unsure; choose a link only when you intentionally maintain one shared source.

  1. Import frequently used skills into PromptHub.
  2. Tag each skill by target tool and purpose.
  3. Use links for stable, actively maintained project skills.
  4. Use copies for temporary distribution or movement between machines.
  5. Recheck installed targets after changing an important skill.

Use Skills for durable behavior rules and tool capabilities; use Prompts for specific model-call templates.

Use this workflow for shared skills:

  1. Open the detail page and confirm the skill solves your problem.
  2. Copy it into your own skill library.
  3. Replace other people’s paths, tools, team rules, and wording.
  4. Remove permissions, external services, or risky steps you do not need.
  5. Test it in a low-risk project.
  6. Install it into your daily AI tool only after it behaves well.

If a skill tells the AI to read files, run commands, upload content, or call external services, review it carefully first.

Local skills are your private capability library. Community skills are public entries other people can discover.

Before using a community skill, check:

  • Which AI tool it targets.
  • Whether it is for coding, documentation, release checks, or another workflow.
  • Whether the instructions are clear.
  • Whether you need to adapt it to your own team rules before installing it.

Before publishing your own skill, check that SKILL.md is understandable on its own, names supported tools, explains input and output, and does not expose private paths, customer names, repository URLs, secrets, or internal team rules.

See Publish to Community for the full review flow.

A Prompt is a template for a model call. A Skill is a reusable capability installed into an AI tool and may include supporting files. Mature workflows often use both: prompts for specific tasks and skills for durable behavior or tool procedures.