Guide · Agent skill
How to use the implement agent skill: spec and ticket workflow
Updated · written with AI from the project's own documentation and checked against our catalogue data
The implement skill, from mattpocock/skills under skills/engineering/implement (MIT licence), tells an agent how to build a piece of work described in a spec or a set of tickets. It follows a fixed sequence of fetching the ticket, building test-first, checking types and tests, reviewing, and committing. This guide explains what the skill does, how to install it, how to trigger it, and what it needs before you rely on it.
Implement From Specmattpocock/skills · MIT Download the skill Quick start
- 1Check your requirements. Confirm you have an agent runtime that supports SKILL.md skills, a git repository with a current branch, and an issue tracker if you plan to use ticket references.
- 2Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ inside your project.
- 3Or upload it in the Claude apps. In the Claude apps, go to Settings, then Capabilities, then Skills, and upload the skill.
- 4Make the companion skills available. Ensure the 'tdd' and 'code-review' skills are present, since the implement workflow calls them by name.
- 5Switch to a suitable branch. Check out the branch you want commits on, ideally a feature branch rather than a shared main branch.
- 6Make the request. Ask the agent to implement a named spec or ticket reference, for example ticket 482 from the issue tracker.
- 7Review the output. Check the ticket title the agent stated, the test and typecheck results, the code review, and the commit it made.
What is an agent skill?
An agent skill is a folder containing a SKILL.md file. The file has a short header with a name and description, followed by instructions the agent follows when the skill is in use. Skills let you give an agent a repeatable workflow instead of re-explaining it in every chat.
The implement skill is one of these folders. Its main file is SKILL.md, and the package also includes agents/openai.yaml.
What the implement skill makes the agent do
When you ask the agent to implement a spec or a set of tickets, the skill directs it through a fixed workflow rather than letting it start coding immediately.
- If you pass a ticket reference, the agent fetches it from your issue tracker and states the ticket title before starting. If the reference is ambiguous, it asks you.
- It calls the 'tdd' skill where possible, at pre-agreed seams, to build the work test-first.
- It runs typechecking regularly, runs single test files regularly, and runs the full test suite once at the end.
- When the work is done, it calls the 'code-review' skill to review the changes.
- It commits the work to the current git branch.
How to install the implement skill
Download the ZIP for this skill from this page. Unzip it into one of these locations:
For a personal skill available across your projects, unzip it into ~/.claude/skills/.
For a skill that only applies to one project, unzip it into the .claude/skills/ folder inside that project.
Alternatively, in the Claude apps, open Settings, then Capabilities, then Skills, and upload the skill there.
- Personal: ~/.claude/skills/
- Project: .claude/skills/ inside your repository
- Claude apps: Settings → Capabilities → Skills
How to trigger the skill with natural requests
The SKILL.md header sets disable-model-invocation to true. This means the agent is not meant to load the skill on its own for every task, so you should make the request clearly. Name the spec or ticket you want built, for example: implement the spec in this file, or implement ticket ABC-123 from our tracker.
Include the ticket reference when you have one. The skill fetches it and confirms the title before any code is written, which lets you catch a wrong ticket early.
- Implement the spec in docs/export-spec.md and commit to my branch.
- Implement ticket 482 from the issue tracker.
- Implement the three tickets I pasted below, one at a time.
What the skill needs before you use it
You need an agent runtime that supports SKILL.md skills, and a git repository with a current branch for the commits to land on.
You need access to an issue tracker if you want ticket references resolved. The files do not specify which tracker is supported or how credentials are provided, so check your own setup for that.
The skill refers to two companion skills by name: 'tdd' and 'code-review'. Make sure those are available to your agent, or the workflow's test-driven and review steps will not run as described.
- An agent runtime that supports SKILL.md skills
- A git repository with a current branch
- Access to an issue tracker for ticket references
- The companion skills 'tdd' and 'code-review'
Safety notes: network access and commits
The skill uses the network, because it fetches tickets from your issue tracker. It does not run any scripts that ship with it.
The skill commits to your current branch without asking. If that branch is shared, such as main with no review process, consider switching to a feature branch before you start.
- Network: yes, for issue tracker lookups
- Scripts: none included
- Commits: made directly to the current branch, without confirmation
Tips for getting good results
Write specs or tickets with clear, testable acceptance criteria. The skill builds and tests against what you describe, so vague input produces vague results.
Give a single ticket reference when you can. If a reference is ambiguous, the agent is meant to ask rather than guess, so answer those questions before it begins.
- Check the title the agent states after fetching a ticket.
- Use a dedicated branch for experiments you may want to discard.
- Read the commit before pushing it.
What people use it for
- Building a feature from a written spec file in a repository.
- Working through a batch of tickets from an issue tracker, one reference at a time.
- Adding a change with tests and a review pass before it is committed to a feature branch.
- Keeping a repeatable build, test, review and commit routine across a team that uses the same skill.
Frequently asked questions
Is the implement skill free to use?
The skill is published under the MIT licence, as stated on this page. Check the licence file in the source repository for the full terms before you use or redistribute it.
What do I need to install the implement skill?
You need an agent runtime that supports SKILL.md skills. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings, Capabilities, Skills in the Claude apps.
Does the implement skill work without an issue tracker?
The skill can still implement a spec or work described directly to the agent. Ticket references need access to an issue tracker, and the source does not specify which tracker types are supported.
Why does the implement skill commit without asking me?
The skill instructs the agent to commit its work to the current branch, with no confirmation step in the source. Use a feature branch, not a shared main branch with no review, to limit the impact.
What if the agent cannot find the tdd or code-review skill?
The implement workflow calls those skills by name, so they must be installed and available to your agent. Without them, the test-driven and review steps described in the workflow will not run as written.
Is the implement skill good for quick prototypes?
It is not a good fit for exploratory prototyping or spikes, because the workflow forces tests, a review and a commit. Use a plain agent conversation when you do not want those steps yet.
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 the implement skill if you already write specs or tickets and want a consistent build, test, review and commit loop, especially with a connected issue tracker. Pick something else for exploratory prototyping, or for repositories where committing straight to a shared branch without review is unsafe.
Download the skill