Skills for AI Agents

Featurevisor skills are capabilities for AI agents, that you can install and use with your favourite tools.

The /featurevisor skill teaches your agent to:

• Author features, segments, and attributes
• Write complex conditions and rollout rules
• Generate test specs and run them
• Answer questions like "what's enabled where?" and "why is checkout disabled for userId 123 in NL?"
• Debug evaluations without you having to read the definitions yourself

Installing skills#

Via skills.sh you can install skills running:

$ npx skills add featurevisor/featurevisor

You will be prompted to choose which agent(s) you want to install the skill for, including:

• Claude Code
• Codex
• Cursor
• OpenCode
• ...and more

Using skills#

Once installed, you can use the skill by typing /featurevisor in your chat.

Several examples below:

Create a feature with a gradual rollout#

/featurevisor create a showWishlist feature, tagged web, starting at 5% in production

Spin up an A/B test#

/featurevisor add an A/B test on the pricing page - control vs treatment, 50/50, only in Germany and the Netherlands

Remote configuration#

/featurevisor expose a checkoutConfig feature with paymentMethods (array) and allowDiscountCode (boolean), override paymentMethods to [paypal, ideal] for NL

Write a complex targeting segment#

/featurevisor create a segment for adult iPhone users in NL or DE who are not on the newsletter

Set up mutually exclusive experiments#

/featurevisor put oneClickCheckout and expressShipping in a mutex group, 50/50 slots

Force-enable for QA in production#

/featurevisor let the QA team see wishlist in production while it’s still at 0% for everyone else

Answer "what's enabled where?"#

/featurevisor list every feature enabled in production that has an A/B test running and no test spec

Debug why an evaluation went the way it did#

/featurevisor why is checkout disabled for userId 123 in NL?

Generate test specs for an existing feature#

/featurevisor write tests for sidebar covering all variations, all environments, and the QA force rule

Find and clean up dead weight#

/featurevisor find unused segments and attributes, plus any duplicate segments

Deprecate a feature#

/featurevisor deprecate the oldCheckout feature with a 2-week grace period

Black Friday segment#

/featurevisor add a Black Friday segment and ramp showPromoBanner to 100% in it

Simulate traffic distribution#

/featurevisor simulate the 25% wishlist rollout against 10k users — will it actually hit 25%?

Explain a feature#

/featurevisor explain the feature checkoutRedesign

Source#

The content of the skills is open source and available on GitHub.