Arkwright Factory — Git skill for Claude Code
Prompt-defined agents that take software work from intake to pull request, for Claude Code, Codex CLI, and GitHub Copilot CLI.
How to install Arkwright Factory
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open matt-major/hugo-factory and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Arkwright Factory does
Prompt-defined agents that take software work from intake to pull request, for Claude Code, Codex CLI, and GitHub Copilot CLI.
Alternatives in Git
- Worktree Deliver — Commit all work, push, and create a pull request from the current worktree 23.4k ★
- Yeet — Stage, commit, push code, and open a GitHub pull request via CLI 14.6k ★
- 40+ Claude Code Tips — Tips for getting the most out of Claude Code, including a custom status line script, cutting the system prompt 5.8k ★
README
Arkwright Factory
Arkwright is a team of coding agents that takes a piece of work from a rough request to a reviewed pull request. You describe a bug or a feature to Arkwright. It investigates, writes a plan for you to approve, has the change made and reviewed, and hands you a PR to merge.
It works with [Claude Code](https://code.claude.com), [Codex CLI](https://github.com/openai/codex), and [GitHub Copilot CLI](https://github.com/github/copilot-cli). The agents are plain prompt files, so there is nothing to build or run apart from the tool you already use.
How it works
You only ever talk to Arkwright. Behind it are four specialist agents:
| Agent | Job |
|---|---|
arkwright |
Talks to you and hands work to the other agents. Never edits code. |
triage |
Researches the request, reproduces bugs, and finds or creates the issue. |
plan |
Interviews you, then writes the plan and commits it to the branch. |
implement |
Makes the change, proves it works, and opens the PR. |
review |
Reviews the PR adversarially and reports back to Arkwright. Never edits code. |
flowchart LR
req([Your request]) --> triage
subgraph arkwright [Arkwright coordinates]
direction LR
triage[Triage] --> plan[Plan]
plan --> approve{You approve?}
approve -->|yes| implement[Implement]
approve -->|changes| plan
implement --> review[Review]
review -->|fixes needed| implement
end
review -->|accepted| merge([You merge the PR])
classDef user fill:#fde68a,stroke:#b45309,color:#451a03
classDef agent fill:#bfdbfe,stroke:#1d4ed8,color:#172554
classDef gate fill:#fecaca,stroke:#b91c1c,color:#450a0a
classDef done fill:#bbf7d0,stroke:#15803d,color:#052e16
class req user
class triage,plan,implement,review agent
class approve gate
class merge done
Arkwright skips stages that aren't needed: a trivial fix doesn't need a plan, and a link to an open issue doesn't
Related Skills
Actual Factory Demo
Build a 5-agent software factory in 15 minutes, and take an example ASCII art task from task definition to Pul
The 3 Clis
Three agent-CLI harnesses (GitHub Copilot CLI, OpenAI Codex CLI, Claude Code) that take a framework pack over
Instant Factory
One prompt that turns a GitHub repo into a Claude Code routine software factory
Busybees
A software factory of Claude Code sessions driven by GitHub issues, labels and pull requests
Claude Amit Mizrachi
Claude Code skills for unattended work: hand a session off (next-prompt), take a feature through the night as
Open Contribute
Agent Skills that take you from "I want to contribute to open source" to a submitted pull request — and measur
Related Agents
Developer Overview
code-container (container) creates isolated Docker environments for AI coding harnesses (Claude Code, OpenCode
CLI Engineer
Ingénieur du CLI Hookstack (packages/cli/, package npm hookstack-cli). Utiliser pour toute évolution du CLI in
Sf ML Engineer
ML/DL Engineer of the Software Factory cell. Use for machine learning and deep learning work - data contracts,