Guide · Agent skill
How to use the Verify Changes Before Claiming Done agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The verify skill is an agent skill for coding assistants that stops the agent from declaring a feature, fix, or refactor finished without evidence. It is written for Claude-style agent skills, so it works when you add it to Claude Code or the Claude apps. After reading this guide you will know what the skill makes the agent do, how to install it, how to trigger it, and what to expect in its report.
Verify Changes Before Claiming DoneYeachan-Heo/oh-my-claudecode · MIT Download the skill Quick start
- 1Download the ZIP. Get the skill ZIP for skills/verify from the source page.
- 2Unzip into a skills folder. Extract it into ~/.claude/skills/ for personal use, or into .claude/skills/ inside a project.
- 3Or upload in the Claude apps. Open Settings, go to Capabilities, then upload the skill under Skills.
- 4Name the behavior to prove. Ask the agent to verify a specific behavior, such as a fix or refactor, before it says the work is done.
- 5Review the evidence report. Check what was verified, which commands ran, what passed, and what failed or remains unverified.
What the verify skill makes the agent do
Agent skills are instruction packages that tell an agent how to handle a certain kind of task. The verify skill covers the end of a coding task. Its goal is to turn vague statements like "it should work" into concrete evidence.
When the skill applies, the agent first identifies the exact behavior that must be proven. It then works through a fixed verification order: existing tests first, then typecheck or build, then the narrowest direct command checks available, and finally manual or interactive validation if automation is not enough.
- Identify the exact behavior that must be proven.
- Prefer existing tests first.
- If coverage is missing, run the narrowest direct verification commands available.
- If direct automation is not enough, describe manual validation steps and gather observable evidence.
- Report only what was actually verified.
Install the verify skill
Download the ZIP for skills/verify from the source page. The skill ships as a single file, SKILL.md. Unzip it so the skill folder ends up in one of these locations:
You can install it for your user account by unzipping into ~/.claude/skills/. You can also install it for a single project by unzipping into the .claude/skills/ folder inside that project.
If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload
How to trigger the verify skill with natural requests
The skill is meant to be used when you want confidence that a feature, fix, or refactor actually works. You do not need a special command. Ask in plain language that makes the goal clear, for example by asking the agent to prove the change works or to verify before it wraps up.
The skill's description is "Verify that a change really works before you claim completion," so requests that match that intent are the ones it is designed for.
- "Verify that the login fix really works before you call it done."
- "Prove the refactor didn't change behavior. Run the existing tests first."
- "Before you say this feature is finished, show me what you actually checked."
What the agent reports back
The skill asks for a short evidence report rather than a wall of logs. The report covers what was verified, which commands or tests were run, what passed, and what failed or remains unverified.
If a check fails, the failure should be shown clearly. If no realistic verification path exists, the agent is supposed to say so directly instead of claiming success.
- What was verified
- Which commands or tests were run
- What passed
- What failed or remains unverified
Requirements and safety notes
The skill lists no specific requirements. It does not run scripts and does not use the network, according to its listing. Its only file is SKILL.md.
The practical limit is the project itself. The skill can only verify what your project makes possible. If there are no tests, no build tooling, and no runnable commands, the agent can only report that verification is not possible.
- Runs scripts: no
- Uses network: no
- Files: SKILL.md only
- Projects with no tests or runnable commands get little value from it
Tips for getting useful results
Name the behavior you care about. The skill starts by identifying what must be proven, so a specific request produces a more useful check than a general one.
Keep your existing test suite healthy. The skill prefers existing tests, so the better your tests, the stronger the evidence.
- Name the exact behavior, such as a specific input and expected output.
- Mention the test command your project uses if you know it.
- Expect the agent to add a required evidence step, so it is slower than a quick prototype.
- Read the unverified section of the report; it tells you what still needs a human check.
What people use it for
- Confirming an AI-written bug fix runs through the existing test suite before merging.
- Checking that a refactor did not change behavior by running the project's current tests.
- Getting a clear list of failing checks instead of a vague success message.
- Documenting which parts of a change still need manual testing.
- Keeping verification consistent across a team that relies on the same test suite.
Frequently asked questions
How do I install the verify agent skill?
Download the ZIP for skills/verify from the source page and unzip it into ~/.claude/skills/ for personal use or into .claude/skills/ inside a project. In the Claude apps, you can instead upload it under Settings, then Capabilities, then Skills.
Does the verify skill require anything else to run?
The listing shows no listed requirements. The skill does not run scripts and does not use the network. Its real dependency is your project, since it needs tests, build tooling, or runnable commands to verify anything.
Is the verify skill free?
The skill comes from the oh-my-claudecode project and is licensed under MIT. The source material does not list a price, so check the source page for current terms.
What should I do if the skill says verification is not possible?
That means no realistic check was found for your project. Add tests or a runnable command, or follow the manual validation steps the agent describes and gather observable evidence yourself.
Does the verify skill run my code or send data anywhere?
According to its listing, it runs no scripts and uses no network. It is a single SKILL.md instruction file. Any commands the agent runs during verification are the ones your project already supports.
Who should not use the verify skill?
Skip it for quick exploratory prototyping where verification is not needed yet, since it adds a required evidence step. It also adds little value in projects with no tests, build tooling, or runnable commands.
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 verify skill if you want evidence that AI-written changes work before you merge, especially if your project has a test suite the agent can run. Skip it for throwaway prototypes or projects with no tests or runnable commands, where it can only report that verification is not possible.
Download the skill