Ohmyharness — DevOps skill for Claude Code

omh — the safe way to run an agent against your repo

A sandboxed session, your setup already in it, and a branch you can review.

[](https://github.com/mindsers/ohmyharness/actions/workflows/ci.yml) [](https://github.com/mindsers/ohmyharness/releases) [](LICENSE)

$ omh init              # detects your stack, decides, reports. no questions.
$ omh new claude        # sandboxed, curated, your setup already inside
$ omh sessions attach   # open that same session in your editor
$ omh graph             # browse your codebase as a graph

**Status: early, and one harness deep.** `0.8.0`. **Claude Code is the only harness anyone has done real work through.** `opencode` and `omp` pass `omh doctor`, which proves their paths are right and nothing whatever about their behaviour — so *declare once, switch harness* is the shape the architecture is built for and not yet a thing anybody has done.

What is verified is the loop below: a sandboxed session with your config already inside it, and a branch you can read before it touches your checkout. This release closes the two silent failures that would have got worse with every user — two checkouts of the same name no longer share sessions, and the carried-file scan now says what it could not read — and adds [`omh eject`](docs/commands.md#omh-eject-harness---to-dir), which writes the config out and steps aside.

[What isn't done](#what-isnt-done) is a real list, not a modesty ritual.

---

What you actually get

Running `omh new claude` instead of `claude` buys five things:

**A sandbox that protects your repo, not just your host.** The agent works in a git worktree on its own branch. Your checkout is never mounted. Review with `omh sessions diff`, ship with `omh sessions commit` and `omh sess