Agent Ops — Development skill for Claude Code

Loop Engineering Toolkit

Loop Engineering Toolkit is an evidence-driven development-loop toolkit for agy, Codex, Claude Code, and opencode. It is designed to turn acceptance criteria, explicit verification, safe lifecycle hooks, and independent review into a repeatable engineering workflow.

This repository is in its foundation stage. The CLI, hook runtime, normative specification, and installation profiles are being developed as a reviewable pre-1.0 interface.

The CLI is published as `@kylecheng3146/agent-ops` as a pre-1.0 interface; command behavior may change before 1.0.

Quick start from npm

Requires Node.js `>=22.14.0`. Install the published CLI globally:

npm install --global @kylecheng3146/agent-ops@latest
agent-ops --version

You can run it without a global install with `npx`. When the command runs in an interactive terminal without arguments, it opens the setup wizard automatically:

npx --yes @kylecheng3146/agent-ops@latest

Use `--help` for the complete command reference or provide explicit options in automation. The wizard never writes files until you review and confirm its installation plan.

The interactive multi-select screens start with no harness or profile selected. Choose at least one of `agy`, `codex`, `claude`, and `opencode`, and at least one of the `core`, `advisory`, `guardrails`, and `loop` profiles before confirming. For scripted use, `--harness all` selects all four harnesses; comma-separated selections such as `codex,opencode` are also supported. The legacy `both` value remains an alias for `codex,claude`.

Preview a project installation before changing files:

agent-ops init \
  --dry-run --scope project --harness all --profile core --json

After reviewing the plan, apply it explicitly with `--yes`:

agent-ops init --scope project --harness all --profile core --yes

The remaining day-to-day checks are:

agent-ops trust status --json
agent-ops doctor --json
agent-ops config e