Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →withastro/changeset
Create a changeset for the Astro monorepo.
Passed 21 of 24 test cases with the skill, and 14 of 24 without it.
and 6 more in the full report
7 fixed14 passed either way3 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
We are removing the deprecated `legacyOutput` configuration key in `astro` to streamline build targets. Write a proper changeset description block for this PR that explains the breaking change and provides a clear migration snippet for users upgrading their site configs.
| Checked | Without skill | With skill |
|---|---|---|
| Sets the semver bump type to major. | Passed | Passed |
| Avoids level 2 or level 3 headings in the changeset text. | Missed | Passed |
| Begins the description with a present tense verb. | Missed | Passed |
The instructions your model reads when the skill loads.
Create changeset files for the Astro monorepo. Changesets declare which packages changed, the semver bump type, and a user-facing message that becomes the CHANGELOG entry.
Every PR that modifies a package requires a changeset. Only examples/* changes are exempt.
Run pnpm changeset --empty from the repo root. This creates a randomly-named .md file in .changeset/ with empty front matter — no need to invent a filename or inspect the directory. Then edit the generated file to add the package bump and message.
---
'<package-name>': patch
---
<changeset message>
name field in the package's package.json exactly (e.g., 'astro', '@astrojs/node')patch, minor, or majormajor and minor bumps to the core astro package are blocked by CI and require maintainer reviewThe changeset message is a public CHANGELOG entry. Write it for Astro users, not for code reviewers.
Begin with a present tense verb that completes the sentence "This PR ...":
Describe the change as someone building an Astro site will experience it, not how it was implemented internally:
// Too implementation-focused
Logs helpful errors if content is invalid
// Better -- user-facing impact
test cases passed, vs 14 / 24 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.