Guide · Agent skill
How to use the Build and Test Runner agent skill (build-test)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The build-test skill is a reusable instruction package for an AI coding agent. After the agent edits code, it runs your project's own build, typecheck and test commands and records whether they passed. This guide is for developers who have not used agent skills before; by the end you will know what the skill does, how to install it, how to trigger it, and what to check before running it.
Build and Test Runnernexu-io/open-design · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Get the build-test ZIP from this page. It contains open-design.json and SKILL.md.
- 2Install the skill. Unzip into ~/.claude/skills/ for personal use or into .claude/skills/ in your project. Alternatively, upload the ZIP under Settings → Capabilities → Skills.
- 3Confirm prerequisites. Make sure Node.js and pnpm (or your project's package manager) are installed, and that the open-design daemon runner is available.
- 4Check your scripts. For custom stacks, confirm package.json has typecheck and test scripts, since the skill reads them from there.
- 5Make your edits. Ask the agent to change code as usual, using the patch-edit step or your normal workflow.
- 6Ask for verification. Request that the build and tests be run after the edits. The skill executes the commands and writes the report and log.
- 7Review the results. Open critique/build-test.json and critique/build-test.log to see which commands ran, how long they took and any failures.
What the build-test skill makes the agent do
Instead of relying only on a subjective review of the code, the agent runs the project's declared typecheck and test commands after it edits code. For the Next, Vite and Remix frameworks the defaults are pnpm typecheck and pnpm test. For custom stacks, the commands are read from the scripts section of package.json.
The skill records whether the build and tests passed, how long they took, the commands it ran, any failures, and the raw output in a log file. Results are written to critique/build-test.json and critique/build-test.log.
- build: passing or failing, based on the typecheck command exiting with code 0
- tests: passing, failing or skipped, based on the test command exiting with code 0
- A skipped test result must include a non-empty reason
- The log is truncated to the runner's logBudgetBytes limit, which defaults to 1 MiB
Install the build-test skill
Download the ZIP for this skill from this page. The package contains open-design.json and SKILL.md. You can install it in one of two ways.
For a personal install, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- Personal: ~/.claude/skills/
- Project only: .claude/skills/ inside your repository
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
~/.claude/skills/Check the requirements before you run it
The skill needs Node.js and pnpm, or the package manager your project declares. It also needs the open-design daemon runner to execute the atom. For custom stacks, your package.json must include typecheck and test scripts, because the skill reads those commands rather than guessing them.
The skill runs scripts and does not use the network, according to its listing.
- Node.js and pnpm (or the project's declared package manager)
- The open-design daemon runner
- typecheck and test scripts in package.json for custom stacks
Trigger the skill with natural requests
Ask the agent in plain language after it has made changes. Name the outcome you want, such as confirming the build still works after a migration. The agent should load the skill when the request matches its description of running build, typecheck, lint and test commands.
You can also override the test command for a single run. In the open-design plugin pipeline this is done with an input flag, for example testCommand set to pnpm test.
- "Run the build and tests to confirm the migration didn't break anything."
- "Typecheck and run the test suite after your edits, and tell me if anything fails."
- "Verify the build still passes after the refactor."
od plugin run --input testCommand='pnpm test'Use it in a repeating verify loop
In the open-design pipeline, the skill is usually placed after a patch-edit step inside a devloop. The loop repeats until both the build and tests pass or an iteration limit is reached. The example limit in the source is 8 iterations.
The skill only runs when the current plan step is in the completed state, so it does not test a half-edited working tree. It also must not make tests pass by changing test files.
- Build signal: build.passing is true when typecheck exits 0
- Test signal: tests.passing is true when the test command exits 0
- Legacy critique.score is 5 when both pass, 3 when only build passes, and 1 when both fail
{"id":"verify","atoms":["patch-edit","build-test"],"repeat":true,"until":"(build.passing && tests.passing) || iterations >= 8"}What people use it for
- Confirming a framework migration, such as moving a Next, Vite or Remix project to a new version, still typechecks and passes tests.
- Running an automated refactor inside a devloop that keeps fixing code until the checks succeed or 8 iterations are reached.
- Producing a stored build and test report with timings and captured output as evidence after a code change.
- Checking a custom stack whose typecheck and test scripts are defined in package.json.
Frequently asked questions
How do I install the build-test agent skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
What do I need to run the build-test skill?
You need Node.js and pnpm, or the package manager your project declares, plus the open-design daemon runner that executes the atom. Custom stacks also need typecheck and test scripts in package.json.
Is the build-test skill free?
This page does not state pricing. The skill is listed under the Apache-2.0 licence, which you should read in the source repository to confirm the terms of use.
Does the skill need an internet connection?
According to its listing, the skill does not use the network, but it does run scripts. Because it executes project commands, only run it on repositories you trust.
Why does the skill say tests failed when I expected them to pass?
The skill only checks exit codes from your typecheck and test commands and records the output in the log file. Open critique/build-test.log to see the captured stdout and stderr and find the failing step.
Can the build-test skill judge code style or design quality?
No. It only reports whether the configured build, typecheck and test commands exit successfully. Use a linter, a code review or a separate design review for style and quality questions.
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 build-test if you run code migrations or automated refactors and need proof that the build and tests still pass, especially in Next, Vite or Remix projects with defined test commands. Pick something else if your project has no typecheck or test script, or if you want the agent to judge design quality or code style.
Download the skill