accelerate-test
A/B test lifecycle — plan, create, monitor, review, or end a test. What should I test? Set up a split test. How is my test doing? Is there a winner? Stop or declare a winner. NOT for composing a single on-brand variant without a test (use accelerate-design) or running a multi-round optimization loop (use accelerate-evolve).
Install
With your agent
Install OMGSkills skill "humanmade/accelerate-ai-toolkit:skills/accelerate-test" into this agent's normal skills directory: call the OMGSkills MCP get_install_plan tool for your agent and run the script it returns (it installs only the pinned commit and verifies SKILL.md), and stop if MCP or the pin is unavailable. Do not overwrite or change existing skills without asking me.
Manual command
git clone https://github.com/humanmade/accelerate-ai-toolkit /tmp/accelerate-ai-toolkit && ln -s /tmp/accelerate-ai-toolkit/skills/accelerate-test ~/.claude/skills/accelerate-testFrom README
Accelerate — A/B testing lifecycle You handle the full lifecycle of A/B tests on an Accelerate site: planning, creating, monitoring, reviewing, and ending them. Figure out which phase the user is in from context, then run the appropriate flow. Every factual statement in the user-facing response must be supported by the current readings. Do not invent a test name, version name or index, or choose one explanation when the evidence allows several. Zero or missing impressions do not determine conversions or other measurements: if impressions are the only supplied count, describe only impressions, not "no measurements", "no results", or "no data". Missing fields stay unknown. When versions remain after completion, say "I can't verify that the winner is saved as the block's content"; do not shorten this to "the winner isn't applied" or "isn't in place." Check the response itself against these rules, not only the intended actions listed after it.