Overview
agent-harness is an open-source workflow skill for Claude Code, published by alirezarezvani. Its author describes it as: “Turn any domain folder of skills into a bounded agentic loop: compile a goal into a verifiable task plan, execute tasks with the domain's own tools, verify every task with machine-run checks, retry with caps, escalate…”. The project has 25k stars on GitHub and is available under the MIT license. Add it to your setup with `/plugin marketplace add alirezarezvani/claude-skills`.
What agent-harness does
You are a harness operator, not a hero. The loop — not your optimism — decides when work is done. Your job: compile the goal into tasks with checks, execute one task at a time, let the controller adjudicate verification, and stop when the state machine says stop.
What’s inside
The SKILL.md for agent-harness is organised into these sections:
- The contract
- Quick start
- Hard rules
- Exit codes (branch on these mechanically)
- Verifiable success
- Related skills
Installation
Add agent-harness to your agent with:
/plugin marketplace add alirezarezvani/claude-skillsReview a skill’s source before installing it — skills can instruct your agent to run commands. This command comes from the skill’s public repository, which is the source of truth for setup steps. New to skills? Read how to install Claude Code skills.
Supported agents
- Claude Code — named in the skill.md
- Codex — declared in the repository topics
- Cursor — declared in the repository topics
- Gemini CLI — declared in the repository topics
License
agent-harness is published under the MIT license.
Strengths
- Clear MIT license — safe to read and adapt within its terms
- Ships in alirezarezvani/claude-skills, an established project with 24,617 GitHub stars
- Works with Claude Code, Codex, Cursor, Gemini CLI
- Actively maintained — repository updated in the last 90 days
Author
Topics
Frequently asked questions
- What does agent-harness do?
- Turn any domain folder of skills into a bounded agentic loop: compile a goal into a verifiable task plan, execute tasks with the domain's own tools, verify every task with machine-run checks, retry with caps, escalate to a human when budgets exhaust, and refuse to close until everything is verified or explicitly waived. Use when you want an agent or subagent to pick up a goal and drive it to a verified close across one of this repo's 18 domains ('run this goal through the engineering harness', 'set up an agentic loop for marketing work', 'make the finance domain self-verifying'). NOT for authoring Claude Code Workflow-tool .js scripts (workflow-builder), N-agent tournaments on one task (agenthub), single-file metric optimization (autoresearch-agent), or discovering published loop recipes (loop-library).
- How do I install agent-harness?
- Run /plugin marketplace add alirezarezvani/claude-skills in your agent, then reload your skills. Review the source at https://github.com/alirezarezvani/claude-skills before installing.
- Is agent-harness free to use?
- Yes. agent-harness is free and open source under the MIT license, so you can read, run, and adapt it within that license's terms.
- Which agents does agent-harness work with?
- agent-harness lists Claude Code, Codex, Cursor, and Gemini CLI — named in its SKILL.md, its plugin marketplace or its repository topics. Skills use the portable SKILL.md format, so other agents that read that format may load it too.
- Where does agent-harness come from?
- agent-harness ships inside alirezarezvani/claude-skills, a repository that contains 41 catalogued skills in total. The repository's 24,617 GitHub stars apply to that whole collection, not to this skill on its own.
