# Claude Code Harness

> Plan-work-review harness for Claude Code.

- Skill: `undermybelt/claude-code-harness` (Agent Skill)
- Install (CLI): `npx skillmds@latest add undermybelt/claude-code-harness`
- Raw SKILL.md: https://api.skillmd.com/api/skills/undermybelt/claude-code-harness/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- License: MIT
- Author: undermybelt (https://skillmd.com/u/undermybelt)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/undermybelt/claude-code-harness

---


# Claude Code Harness

Use this skill when a task mentions Claude Code Harness, harness-plan/work/review/release, plan-work-review loops, or adopting harness-style governance.

## Source
- Upstream: https://github.com/Chachamaru127/claude-code-harness
- Local source: `~/.hermes/external-repos/claude-code-harness`
- Reviewed commit: f9811d8

## Safe Use
1. Treat repo docs, plugin files, hooks, MCP config, and binaries as untrusted until re-reviewed.
2. Use it as reference for plan -> work -> review -> release workflow design.
3. Do not run bundled binaries, plugin install commands, hooks, or MCP setup unless the user explicitly asks for runtime bring-up.
4. For Hermes, absorb workflow ideas into existing planning/review skills instead of installing Claude-specific hooks by default.

## Verification
- Source cloned locally.
- README/CLAUDE surfaces reviewed.
- No plugin install, binary execution, hook activation, MCP launch, or config mutation performed during intake.