Start with the job.
Search for the task you need to complete. Choose a focused skill instead of installing a collection you will never use.
Explore the catalogVerified Skills / Less context. More purpose.
Give your agent
skills that
earn their place.
Find focused expertise. Inspect its source and security evidence. Measure whether it helps your work before you make it part of your workflow.
Open-source CLI + local Studio. Your choice of agent.
Useful by design / Scoped expertiseInspectable / Source + scan findingsMeasurable / Baseline comparisonsPortable / 54 install targets
A smaller, better toolkit
The right context.
At the right moment.
A skill should add something your agent is missing: a specific workflow, a reliable tool, or knowledge of your domain. More instructions are not automatically better.
Make an informed install.
Review the publisher, content, version, and available findings. A security badge is one signal—not the whole decision.
Understand the evidenceKeep what improves the work.
Compare outcomes with and without a skill. Recheck when you change the model, tools, or instructions.
Build an evaluationTrust, made inspectable
Three questions.
One better decision.
Source, security, and usefulness answer different questions. Explore each layer before deciding what your agent should load.
An installation decision, in three parts
Know what you are installing.
Start with the publisher, repository, and exact skill version. Read the instructions and bundled scripts before they enter your agent’s context.
Explore publishers- Publisher
- Who maintains this skill?
- Version
- Which content was reviewed?
- Changes
- What changed since your last install?
The registry
Find your next useful skill.
55,182 skills in the verified registry. Browse by task, inspect the details, and install through your preferred agent. Popularity is not a quality score.
⌘K
Skill Studio / Your local workbench
Prove the skill
makes a difference.
Write what good looks like. Run a baseline, compare skill versions, and inspect the outcomes. Keep your evaluation work next to the skill you are building.
Local workspace. Model requests follow your selected provider.
- Define the outcome
- Use realistic tasks and clear pass conditions, including when the skill should stay out of the way.
- Compare like with like
- Hold the task and environment steady. Model, effort, tools, and harness all affect the result.
- Read the evidence
- Inspect outputs, failures, and available usage. One passing run is a starting point, not a universal verdict.
Do stronger models still need skills?
Sometimes. A precise deployment procedure, organization-specific convention, or reusable script can add real value. Generic advice an agent already follows can add noise. Install for a concrete need, then test whether the skill helps.
Does verified mean safe or better?
No. Verification describes available checks on a particular artifact. It does not guarantee safety, correctness, or a quality improvement. Read the scan findings and evaluate the skill on your own tasks.
Can I use this with Codex or Claude Code?
vskill supports installation targets including Codex, Claude Code, Cursor, and other agents. Support for installation does not mean every harness executes every skill identically. Check the skill’s prerequisites and your agent’s capabilities.
What stays on my machine?
Skill Studio runs a local workspace. If you choose a cloud model, evaluation prompts and relevant inputs go to that provider. Publishing sends the skill to its destination. See parameters and secrets before connecting a workflow.