vibecosystem
A large collection of specialized agents, skills and hooks that turns Claude Code into a self-learning AI software team, covering planning, building, reviewing, testing and monitoring.
vibecosystem installs into Claude Code as a set of specialized agents, skills and hooks that plan, build, review, test and learn from mistakes across a software project, aiming to work with no additional configuration beyond installing it. Its agents cover roles such as security scanning, mutation testing, dependency auditing, incident response, release engineering and documentation, alongside skills for tasks such as document generation, website cloning and frontend or full-stack development patterns.
Later releases add agent-level persistent memory, auto-updating documentation, cross-session memory cleanup, a CLI for statistics, health checks, configuration profiles and a dashboard, a real-time terminal status line, a prompt auto-improver that adds project context to vague requests, and a persistent planning system of plan, progress and context files intended to raise task-completion rates.
It installs through an npm package or a one-line install script, and supports several preset profiles intended to reduce token use.