Discover agent skills for real workflows
Search reusable agent workflows, compare source and audit signals, and inspect the bundle before it becomes part of your setup.
Browse by outcome, audit status, maintenance signal, and install activity.
Pick skills by the work they make repeatable.
The right skill is closer to a dependency than a prompt snippet. It should have a specific job, inspectable source, a current snapshot, and enough audit context to decide whether it belongs in your agent setup.
The skill names the repeatable work, not only a persona.
Inputs, outputs, and boundaries are clear before you apply it.
Source, snapshot, files, and audit result all point to the same use case.
It improves one workflow you already run, not every task in your backlog.
Evaluate a skill before it shapes an agent response
Popularity can help you discover candidates, but it should not decide adoption. Use a repeatable review path so every skill is judged by the same signals.
- 01SearchA short list of skills with matching names and descriptions.
Start with the outcome
Use the job you want done: debugging, writing, browser automation, Kubernetes review, research synthesis, or another concrete workflow.
- 02CompareOne candidate that matches the actual files, not just the pitch.
Check the package behind the promise
Open the detail page and read the source, bundled files, snapshot metadata, categories, and audit status before trusting the summary.
- 03AdoptA reusable workflow you can explain to another operator.
Use the narrowest useful skill
Prefer a skill that handles one repeatable workflow well. Broad assistants are harder to review, test, and replace.
- 04RevisitA maintained skill set that stays understandable over time.
Treat updates like dependency changes
When a skill changes source, snapshot, audit status, or category, inspect it again before letting it shape production work.
Separate useful workflow packages from prompt dumps
Repository, owner, and bundled files are visible.
The source is vague, forked without context, or hard to inspect.
The category describes the primary workflow outcome.
Several unrelated categories are mixed together, or no category fits.
Pass, warning, or failure is easy to interpret from the skill page.
A warning is ignored, or a failed audit is treated like a popularity issue.
The copy explains when to use the skill and what it produces.
It promises expertise without naming inputs, outputs, or limits.
Browse the top 12 skill categories
Categories should help you start with intent. Search by the kind of work you need done, then use the skill detail page to confirm that the bundle really supports it.
Aggregators
Multi-source aggregation and gateway agent workflows.
Art & Culture
Art, culture, creative collections, and media archive agent workflows.
Architecture & Design
Architecture, design systems, CAD, and spatial design agent workflows.
Browser Automation
Browser control, web automation, scraping, and extraction agent workflows.
Biology Medicine and Bioinformatics
Healthcare, biology, medicine, and bioinformatics agent workflows.
Cloud Platforms
Cloud provider, infrastructure, and platform operation agent workflows.
Code Execution
Sandboxed code execution and runtime agent workflows.
Coding Agents
Agentic coding, code review, and engineering-assistant agent workflows.
Command Line
Terminal, shell, and command-line workflow agent workflows.
Communication
Messaging, meetings, email, and collaboration agent workflows.
Conversational AI
Chat, voice assistant, and conversational AI agent workflows.
Cryptography
Cryptography, keys, signing, and secure identity agent workflows.
Questions to answer before adopting a skill
What is an agent skill?
An agent skill is a packaged set of instructions and supporting files that helps an AI agent perform a specific kind of work more consistently, such as debugging, writing, research, or operating a domain-specific workflow.
How should I compare two similar skills?
Compare the stated outcome, source repository, install count, categories, and audit result. The better fit is usually the skill with the narrower workflow and clearer maintenance trail, not simply the one with the broadest description.
What do official and community sources mean?
Official sources identify publishers that Wardn Hub marks as trusted or canonical for that catalog entry. Community sources can still be useful, but they deserve closer inspection before use in production agent workflows.
Why do some skills have warnings or no category?
A warning means the automated audit found something worth reviewing. A missing category usually means the primary use case is ambiguous or the taxonomy does not yet have a defensible bucket for that skill.