Creates and tracks hypothesis-driven experiments following a lab methodology cycle.
npx skills add https://github.com/PatriotAi/ai-lab --skill experiment-patriotai
Please help me install this Agent Skill. Skill: experiment Source: https://github.com/PatriotAi/ai-lab/tree/main/.claude/skills/experiment Command: npx skills add https://github.com/PatriotAi/ai-lab --skill experiment-patriotai
experiments/<slug>/ directory with a README based on the experiment template. - Knowledge Capture: Appends a 3-5 line conclusion to docs/learnings.md and promotes successful experiments into skills/ or automations/. - Use Case: You want to check whether a translation skill speeds up bilingual releases. The Skill creates experiments/translate-uaen-speed/README.md with a hypothesis and metric, then records the measured outcome and final decision. ## Quick Start Ask the assistant to start a new experiment for your hypothesis, for example: start an experiment to test whether the translate-uaen skill speeds up bilingual releases.High-intent search queries and answers about installing and using this skill.
Say something like "start an experiment" or "let's test a hypothesis" and state your idea. The Skill formulates a hypothesis and success metric, then creates an experiments/<slug>/ folder with a README from the experiment template.
It follows the lab methodology cycle: hypothesis, design, launch, measurement, keep-or-remove decision, and conclusion. Each experiment is a minimal vertical slice, not a full system build.
Results and observations go into the experiment's own README.md inside experiments/<slug>/. A 3-5 line conclusion is always added to docs/learnings.md, even when the experiment is removed.
Successful experiments are promoted into the skills/ or automations/ directories. The Skill also updates docs/PLAN.md and the README of the relevant catalog to reflect the promotion.
Avoid it for large features that cannot be reduced to a single vertical slice, since the methodology enforces one experiment per end-to-end slice. It is also unsuitable for work involving secrets or personal data, which are prohibited in experiment notes.
luicabref97Apply PDCA cycles to plan, implement, measure, and act on process improvements.
iadr-devAnalyze session logs to compare planned versus actual outcomes and extract reusable patterns.
marin-communityMaintains append-only Markdown logbooks and coordinating GitHub issues for reproducible experiments.
wanyusheng63-altBootstrap frontend prototypes from a template-based starter codebase.