Guide · Agent skill
How to use the ce-work agent skill: install and run plans
Updated · written with AI from the project's own documentation and checked against our catalogue data
The ce-work skill tells an AI coding agent how to carry out a plan, spec or concrete build request in a git repository: work through the tasks in order, change the code, and run local checks. This guide explains what an agent skill is, how to install ce-work, how to trigger it with natural requests, and what to check before letting it edit and commit your code.
Work Execution CommandEveryInc/compound-engineering-plugin · MIT Download the skill Quick start
- 1Download the ZIP. Download the ce-work ZIP from this page.
- 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use or the project's .claude/skills/ for one repository.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Check prerequisites. Confirm the repository is a writable git checkout and that the reference files are present in the skill folder.
- 5Ask for the build. Give the agent a plan path, spec path or clear build request in plain language.
- 6Review the result. Read the changes and commits, and the code review or blocker report, before you accept the work.
What an agent skill is and what ce-work makes the agent do
An agent skill is a folder containing a SKILL.md file and supporting reference files. When a skill is installed and matches your request, the agent loads its instructions and follows them for that task.
ce-work covers the execution step. Given a plan document, a spec path or a clear build request, it works through the tasks in order, makes the code changes, and runs local checks to confirm they work. It follows bundled reference files that govern how it picks an execution approach, whether to split work across parallel helpers, how it commits, and when to stop and report a blocker.
- Standalone use: after implementation, the skill hands the verified change to a code review and shipping step.
- Return-to-Caller Mode: an outer workflow receives only the implementation and verification result and owns the remaining steps.
Install the ce-work skill
Download the ZIP for ce-work from this page and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.
In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/ce-work
- Project install: .claude/skills/ce-work inside your repository
Trigger ce-work with natural requests
You do not need a special command. Describe the work in plain language and point at the plan. The skill's argument hint accepts a plan path, a work description, or a recovery request that names an existing run id. If you give no input, it uses the latest run.
Use ce-work for implementation from a written plan or a clear build request. For open-ended bugs, the skill says to use ce-debug instead.
- "Implement the plan in docs/plans/billing-export.md."
- "Build the spec at specs/invoice-api.md and run the local tests."
- "Add CSV export to the reports page as described in the request below."
- "Resume the run with the id I give you and report its status."
Know what it needs and what it changes
Before running ce-work, confirm the requirements below. The skill edits files and makes git commits, so review the repository state and the changes it makes.
Do not use it on a repository whose changes you have not reviewed. It is also not the right tool for open-ended bug hunts.
- Git and a writable local repository checkout.
- The bundled reference files in the skill folder. The skill stops if they cannot be read.
- An agent runtime that can read files and run commands, plus the project's own test or build tools as the plan requires.
- Optional: Figma access if the Figma design sync reference is used.
- The skill runs scripts and uses the network.
Safety notes: scripts, network and commits
The skill runs scripts and uses the network. Its reference files describe a controller and external routes, and the agent may send repository content to an external process only when the skill's execution-engine rules select that route. Check which engine a run uses before you start it.
The skill never commits files the user did not offer. Implementation commits name only each unit's owned files. Files that were already dirty when a unit needs them are handled by asking in standalone mode, or returned as blocked in Return-to-Caller Mode.
- Start on a branch, not the real default branch, unless you explicitly direct otherwise in the session.
- Review diffs and commits after each run.
Tips for good results
Give the skill a complete plan. The more tasks, file names and acceptance checks the plan states, the less the agent has to guess. Keep your working tree clean before starting so the skill does not need to ask about pre-existing changes.
If the run stops, it reports a blocker rather than renegotiating the plan. Read the blocker, fix the cause, then ask for recovery with the run id instead of starting over.
- Clean your working tree before starting.
- Name the plan path explicitly.
- Keep the run's run id so you can resume it.
What people use it for
- Implementing a feature from a written plan document, task by task, with local checks after each unit.
- Carrying out a spec at a given path inside an existing git repository.
- Running implementation and local verification inside an outer orchestrator that handles review and shipping itself.
- Resuming an earlier run by its run id to inspect or clean up its state.
Frequently asked questions
How do I install the ce-work agent skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.
What do I need to run ce-work?
You need git and a writable local repository checkout, the bundled reference files in the skill folder, and an agent runtime that can read files and run commands. Project test or build tools are needed if your plan requires them, and Figma access is optional for the design sync reference.
Is ce-work free to use?
The source is released under the MIT licence. This page does not state pricing for the Claude apps or any agent runtime, so check those providers for their own terms.
Does ce-work commit code to my repository?
Yes. The skill edits files and makes git commits, and each implementation commit names only that unit's owned files. Review the repository and the results before you rely on them.
Should I use ce-work or ce-debug for a bug?
Use ce-work when you have a plan, spec or clear build request. The skill says open-ended bugs should go to ce-debug, because ce-work may do the wrong kind of work for an investigation.
What happens if the skill cannot read a reference file?
It stops before the action that file governs and reports which file is missing. It does not rebuild the rules from memory, so restore the missing file in the skill folder and resume.
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 ce-work if you already have a written plan or spec and want an agent to carry out the build with local verification, on a repository you can review. Pick something else for open-ended bug hunts, which the skill itself routes to ce-debug, or if you cannot review the commits and file edits it produces.
Download the skill