Unship is an open-source agent harness and plugin integration framework developed by mbenhard. It is designed to package and orchestrate agent skills specifically for Claude environments and terminal workflows without adding unnecessary runtime overhead. In day-to-day operation, Unship functions as a lightweight skill provider and execution harness. It allows AI models, such as Claude Opus, to execute specific tasks, manage tool selection via hardened pickers, and interact with the local development environment through a structured plugin architecture. Unship distinguishes itself by focusing on minimal dependency packaging. By separating core plugin manifests from heavy test and development dependencies, it ensures that importing skills into Claude agent environments avoids bloated installs and cache pollution. The project is open source on GitHub and under active development with versioned releases focusing on skill hardening and workflow ergonomics.
Claude Plugin Architecture — Standardized manifest and plugin configuration for direct integration with Claude-based agent environments. - Skills CLI Management — Command-line interface support for discovering, configuring, and installing modular agent skills. - Zero-Bloat Packaging — Dedicated lightweight plugin distribution structure to avoid pulling unnecessary devDependencies into plugin caches. - Agent Picker Hardening — Robust tool selection and action invocation mechanisms designed to improve agent execution reliability. - End-to-End Test Harness — Built-in Playwright-based test suite for validating skill interfaces and UI interactions.
Use cases
Use Case: Claude Plugin Skill Integration A developer wants to provide Claude with local tools and skills without bloating their global environment. Unship packages the skills into a standalone manifest that loads cleanly into the agent runtime.
Use Case: Orchestrating CLI Agent Workflows An engineer needs a structured harness to execute terminal-based agent commands and tool actions. Unship provides the runner and tool pickers to guide model execution.
Use Case: Automated Agent Interaction Testing A developer building custom agent extensions needs to verify tool selection and UI interactions. Unship includes end-to-end test fixtures using Playwright to validate behavior before release.
Who it is for
Unship is designed for developers building or consuming Claude plugins and terminal-based agent skills who prioritize lightweight, dependency-free integrations. It is less relevant for teams looking for an all-in-one graphical IDE or turnkey autonomous coding assistants.
Install guide
Installation Clone the repository or install the skill package directly using Git: bash git clone https://github.com/mbenhard/unship.git cd unship npm install
First Run To link or verify the plugin within your Claude environment: bash # Build the plugin artifacts npm run build # Run the test suite to verify skill integrity npm test
Example Session Once configured in your Claude plugin path, interact with the harness through the CLI interface: bash # Execute skill picker validation node src/index.js --help
Claw Code — Claw Code is an open-source terminal-based AI coding assistant developed by Ultraworkers. Built on a hybrid architecture featuring a high-performance…
Ouroboros — Ouroboros is an open-source autonomous CLI coding agent developed by razzant. Built primarily in Python, the agent is engineered to…
g3 — g3 is an open-source, Rust-based command-line coding agent framework developed by Dhanji Prasanna. It is designed to orchestrate LLM-driven software…
OpenHands — OpenHands (formerly OpenDevin) is an open-source autonomous AI software development agent and orchestration platform. It provides an extensible runtime environment…
SWE-agent — SWE-agent is an open-source autonomous software engineering agent harness built and maintained by researchers from Princeton University and Stanford University.…
OpenCode — OpenCode is an open-source AI coding agent developed by Anomaly that operates across terminal environments, desktop applications, and IDE extensions.…