The wshobson/agents repository is a comprehensive agentic plugin marketplace designed to facilitate advanced AI-driven workflows across multiple developer harnesses, including Claude Code, Codex CLI, Cursor, OpenCode, GitHub Copilot, and Gemini CLI. Its primary purpose is to provide production-ready, modular building blocks for agentic automation, enabling developers to leverage a wide array of plugins, agents, skills, and commands tailored to their preferred environments and AI models.
At its core, the repository offers 84 granular plugins, 192 domain-specific agents, 156 modular skills, 102 slash commands, and 16 orchestrators for multi-agent coordination. These components cover a broad spectrum of tasks, such as architecture, programming languages, infrastructure, security, data science, machine learning, documentation, business operations, and SEO. The marketplace is structured so that each plugin is isolated and composable, allowing users to install only the components relevant to their workflow without loading the entire marketplace. This design ensures efficiency and context-specific operation.
A key feature of the repository is its multi-harness support. From a single Markdown source-of-truth, it generates harness-native artifacts for each supported environment. For example, Claude Code serves as the primary source, producing native marketplace and plugin files, while Codex CLI, Cursor, OpenCode, Gemini CLI, and Copilot receive artifacts tailored to their respective formats and requirements. This approach avoids lowest-common-denominator translations, ensuring each harness benefits from idiomatic integration and optimal compatibility. The repository also provides detailed setup instructions and troubleshooting guides for each harness, including specialized documentation for Gemini CLI.
The agentic marketplace employs a tiered model strategy to optimize agent selection based on task complexity and required autonomy. Models such as Fable 5, Opus, Sonnet, and Haiku are mapped to different tiers, ranging from long-horizon autonomous work (like large migrations) to fast operational tasks (such as deployment and content management). This tiered approach allows users to configure agents according to their specific needs, balancing performance and cost.
Quality assurance is a central aspect of the repository. The included plugin-eval framework offers a three-layer evaluation system: static analysis for structural integrity, LLM-based semantic evaluation across multiple dimensions, and Monte Carlo simulations for statistical reliability. This rigorous certification process ensures that plugins and skills meet high standards of functionality and reliability before deployment.
Extensive documentation is provided within the repository, covering plugin catalogs, agent categories, skill descriptions, usage examples, architectural principles, harness capability matrices, authoring guidelines, and quality evaluation methods. The documentation is organized to guide users from basic setup to advanced customization and integration, making it accessible for both novice and experienced developers.
External memory integration is supported via Pensyve, which is included as a git-subdir for Claude Code and maintains direct upstream integrations for other harnesses. This enables persistent memory and enhanced agentic capabilities across all supported environments.
The repository is licensed under MIT, promoting open-source collaboration and contribution. Overall, wshobson/agents stands out as a robust, flexible, and well-documented platform for building, orchestrating, and evaluating agentic workflows in modern AI-powered developer tools.
