Browse the library

A token-frugal multi-agent orchestration plugin for Claude Code that routes tasks between Fable 5, Sonnet 5, and Opus 5.5 to optimize usage.

Box0 is an open-source platform that enables long-running, collaborative, and proactive subagents and agent teams for Claude Code and Codex, with features like persistence, scheduling, and webhooks.

Skills for agentic coding CLIs that store project direction, next tasks, and a definition of done in the repo so the agent works against them and verifies results.

VibesPlug Pro is a universal 147-skill AI agent swarm architecture for agents like Claude Code and Cursor, providing zero-tech-debt development with autonomous full-stack lifecycle management and anti-slop directives.

Gives OpenCode, Claude Code and Codex persistent local memory that records decisions and dead ends and injects relevant context into later prompts.

This GitHub template provides a ready-to-use toolkit for creating and distributing Claude Code plugin marketplaces, including scaffolding, validation, hooks, skills, agents, and CI/CD workflows.

This comprehensive guide offers advanced insights into Claude Code, covering CLAUDE.md architecture, skill authoring, and subagent orchestration with ready-to-use practical examples.

This Claude Code plugin acts as an intelligent harness for development workflows, ensuring changes pass through a "Write Gate" and tracking project history to align Claude with architectural preferences.

This framework enables Pydantic AI agents to delegate tasks to nested subagents, supporting sync, async, or auto execution with token tracking.

This agent plugin provides a TypeScript Harness Workflow Engine for multi-agent systems, enforcing deterministic workflow gates and skills.

skillreaper

thousandflowers

This CLI tool helps AI agents reduce context bloat by identifying and safely pruning unused skills, MCP servers, and subagents.

This provides infrastructure for Antigravity IDE's Orchestrator Agents to manage and delegate tasks to specialized Sub-Agents.

This implements a Parallel-Agent Reinforcement Learning (PARL) paradigm for training AI models to decompose tasks into parallel subtasks and coordinate multiple agents.

This skill integrates AI agent's decisions and reasoning directly into Git history, creating a persistent log of development choices and context across sessions.

This plugin provides a structured "Plan, Go, Close" workflow with quality gates for AI coding sessions, ensuring agents follow plans and verify changes.

An opinionated multi-agent development framework for Claude Code, based on John Carmack's philosophy, tailored with 10 domain experts for a specific greenfield web stack.

An advanced AI agent orchestration framework for Windows/WSL2, utilizing Kether governance and Pi/MCP to manage isolated workers, LSP tools, task tiering, and subagent delegation for coding tasks.

A self-hosted web platform that generates, stores, and shares AI coding rules for 30 different tools, including Claude Code and Cursor, with team blueprints and a CLI.

A multi-agent framework that uses speculative execution and specialized agents to make AI coding CLIs like Claude Code more cost-effective and faster.

Optimizes OpenClaw agent workflows by providing auto-delegation to sub-agents, cost-aware model routing, and clearer task status to prevent main agent stalls and reduce costs.

A 100% AI-managed, curated awesome list of over 400 Korean AI coding agent skills, categorized for easy discovery across Claude Code, Gemini CLI, Codex, and Cursor.

A Claude Code plugin that transforms the AI agent into a full development team (PM→Lead→Dev→QA) with a pre-configured, role-driven workflow, 44 commands, and auto-learning capabilities.

This resource provides a marketplace and specific plugins for Claude Code, offering tools like a selectable writing style and configuration auditing for AI agents.

This tool provides live, deep observability into Claude Code sessions by visualizing model calls, tool usage, subagents, and context window activity from its logs.