agents: Multi-Harness Agent Plugin Marketplace

Key Takeaways#

• agents by wshobson is a multi-harness agent and plugin marketplace resource.
• It targets users of Claude Code, Codex, Cursor, OpenCode, GitHub Copilot, Google Antigravity, and Pi.
• The value is discovery and reusable agent/plugin structure, not a single installable AI model.

What it is#

agents is a GitHub resource that collects agentic plugins and workflows across multiple AI coding harnesses. The description positions it as a multi-harness marketplace for Claude Code, Codex, Cursor, OpenCode, GitHub Copilot, Google Antigravity, and Pi. That makes it useful for developers who are trying to reuse agent patterns across tools instead of rewriting setup from scratch for every assistant.

Why it matters#

The AI coding ecosystem is fragmenting. Each harness has its own vocabulary for skills, agents, commands, plugins, and project rules. A resource like agents helps builders compare those patterns and discover reusable components. Instead of treating every assistant as a separate island, it gives teams a place to inspect how agent roles, task templates, and plugin conventions translate across environments.

How to use it#

Start with one harness you already use, such as Claude Code or Codex, and look for agent definitions that match your real workflow. Good candidates are repeatable jobs: code review, test repair, refactoring, documentation, security checks, migration planning, or release notes. Copy the structure carefully, then adapt names, constraints, and tool permissions to your repo. Avoid installing every agent at once. A smaller set of high-trust agents is easier to evaluate and maintain.

Best fit#

This resource is best for power users and platform teams standardizing AI coding workflows. It is not a model, not a hosted app, and not a complete governance layer. Treat it as a catalog and reference set for agent patterns that still need local review before use in production repositories.