by Hidari
Product-agnostic AI exploratory monkey testing for Claude Code. Wanders a web app with weighted-random actions to surface what humans would notice (console errors, unhandled rejections, 4xx/5xx, error boundaries, layout overflow, broken images, dead-end navigation). Profile-YAML driven with two-layer safety (dispatcher forces production to anonymous read-only + compiles a denylist; explorer matches every action against the denylist at runtime). Markdown + schema only, no hooks/mcpServers (local-first, zero auto-run code).
Claude Code1 Skill