Guide · Agent skill
How to use the software-factory agent skill for factory design
Updated · written with AI from the project's own documentation and checked against our catalogue data
The software-factory skill is a set of shared terms and design rules that an AI agent uses when you plan a system of agents that files, triages, reviews or implements tickets and pull requests. It is for engineers and teams who want a consistent design checklist before choosing tools such as GitHub Actions, Linear or a sandbox provider. After this guide you will know what the skill changes about the agent's behaviour, how to install it, how to ask for it in plain language and what to expect from it.
Software Factory Design Vocabularymattpocock/skills · MIT Download the skill Quick start
- 1Download the ZIP. Get the software-factory ZIP from the page you are reading.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project.
- 3Or upload it in the Claude apps. Go to Settings, then Capabilities, then Skills, and upload the ZIP.
- 4Name the skill in your request. Say you want to use the software-factory skill and describe the problem you want to solve.
- 5Answer the checklist. For each task and ticket source, answer the five questions the skill asks until every service has a product and every task has an answer.
What an agent skill is and what this one does
An agent skill is a folder of instructions and reference files that an agent such as Claude can load when a task calls for it. This one is named software-factory and comes from the mattpocock/skills repository under the MIT licence.
When you ask the agent to plan an automated software factory, it uses a fixed vocabulary: task, trigger, run, runner, ticket tracker, ticket source, git host, orchestrator and observability platform. It steers the design toward safe patterns: start with one throttled automated trigger, keep human review on every pull request, and cap retry loops at a few tries before a human steps in.
- It reads the bundled survey files for each service and then researches current product options itself.
- It walks through a five-question checklist for every task and every ticket source.
- It challenges designs that are more ambitious than the user's problem needs.
- It does not build, run, deploy or open pull requests. It shapes the design conversation.
How to install the software-factory skill
Download the ZIP for the skill from the page you are reading. Then choose one of the following locations.
For a personal install that applies to all your projects, unzip the file into ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project.
If you use the Claude apps instead of the file system, open Settings, go to Capabilities, then Skills, and upload the ZIP there.
- Personal: ~/.claude/skills/
- Single project: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
How to trigger the skill with natural requests
The SKILL.md file sets disable-model-invocation to true, so the agent is not meant to pick this skill up on its own from a casual mention. Name it in your request, for example by saying you want to use the software-factory skill.
The skill works best when you start from a problem rather than a finished architecture. The guide in its bundled examples lists the tasks that solve common problems such as untriaged tickets and pull requests waiting for review. Describe the problem in your own words and let the agent map it onto the vocabulary.
- Use the software-factory skill. Untriaged bug reports pile up and nobody reads them. What should the factory do first?
- Use the software-factory skill. We want an agent to review pull requests. Which tasks and triggers do we need, and what are the human review points?
- Use the software-factory skill. We are choosing between GitHub Actions and a sandbox provider for runners. Walk me through the questions that decide it.
Example: a first design conversation
Suppose your team has a backlog of ready tickets that nobody builds. A useful answer from the skill names one trigger, one task and one human re-entry point, rather than a full factory.
In that design, a human labels a ticket as ready. An implementer run builds the ticket until the automated checks pass and opens a draft pull request. An automated review run then reviews the diff. A human reads the pull request before it merges, and each run ends rather than waiting for them. The skill will ask you to answer the checklist for this task: what triggers it, what it reads and writes, what it triggers next, where a human re-enters, and what stops two runs doing the same work.
What the skill needs and how it stays safe
The skill lists no requirements. It does not run scripts and does not use the network, according to its listing. Its files are reference text.
The design step does ask the agent to research current products for each service. Whether that research is possible depends on what your agent setup can access, so check that before relying on the product comparisons. The survey files say their information may be out of date and products, prices and limits change quickly, so confirm details with each vendor.
- Requirements: none listed
- Runs scripts: no
- Uses network: no in the skill itself; product research depends on your agent's tools
- Product advice may be outdated, so verify it before you commit
What people use it for
- Planning an agent that files and sorts incoming bug reports before a human reads them.
- Designing an automated review run that checks each pull request against the codebase's standards.
- Choosing between GitHub Actions, Linear or a sandbox provider with a consistent checklist.
- Testing an ambitious automation idea with the agent, which is asked to push back on overreach.
Frequently asked questions
Is the software-factory skill free to use?
The skill is published under the MIT licence in the mattpocock/skills repository. Its listing does not mention any cost, but the products it helps you choose may have their own prices.
What do I need before I install the software-factory skill?
The listing lists no requirements. In practice the vocabulary assumes you already have a ticket and git workflow, so the skill is most useful to developers who have one.
Does the software-factory skill build or run an AI agent system for me?
No. It does not build, run, deploy or open pull requests. It guides the design discussion and suggests products for each service.
Why doesn't the agent use the software-factory skill automatically?
The SKILL.md file sets disable-model-invocation to true. Name the skill in your request so the agent knows to use it.
How is this different from choosing tools directly, like GitHub Actions or Linear?
The skill does not replace a product comparison. It gives you shared terms and a checklist first, then asks the agent to research current options for each service slot.
What if the product advice the agent gives is out of date?
The survey files warn that products, prices and limits change fast. Verify current details with each vendor before making a decision.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are an engineer or team planning an AI agent system for tickets and pull requests and want a consistent design checklist and safer defaults. Pick something else if you want a tool that runs code, opens PRs or deploys anything, or if you are not yet working with tickets and git.
Download the skill