agentleFS
iangregsondeviangregsondev/agent-skills/skills/engineering/changeset/SKILL.md

changeset

Skill2 starsChanged 3 months agoMITView on GitHub

Changesets — write the changeset file directly, since the interactive CLI asks questions an agent cannot answer. Use when a change to released code is ready to commit and has no changeset yet, when a check fails with "packages have been changed but no changesets were found", when only docs or tooling changed and the release needs an explicit "nothing to release", or when the user asks for a changeset, a release note, or a version bump. Not for running a release or editing versions and changelogs by hand — automation owns those.

19 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use better-auth-best-practices, turborepo and 5 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • anasahmed07Skill

    better-auth-best-practices

    Skill for integrating Better Auth - the comprehensive TypeScript authentication framework.

    Also in 3 projects that use this

  • VercelSkill

    turborepo

    Turborepo monorepo build system guidance. Triggers on: turbo.json, task pipelines, dependsOn, caching, remote cache, the "turbo" CLI, --filter, --affected, CI optimization, environment variables, internal packages, monorepo structure/best practices, and boundaries.

    Also in 3 projects that use this

  • Microsoft AzureSkill

    humanizer

    Tone enforcement patterns for external-facing community responses

    Also in 5 projects that use this

  • SentrySkill

    deslop

    Strip AI-generated noise from the diff before pushing — extra comments a human wouldn't write, defensive try/catch in trusted code paths, casts to any, inline imports in Python, and other style inconsistencies with the surrounding file. Use this immediately before commit so the diff stays clean.

    Also in 3 projects that use this

  • Google CloudSkill

    triage

    Gather context on a change or bug and plan the implementation.

    Also in 3 projects that use this

  • GoogleSkill

    improve-codebase-architecture

    Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every name must be one Changesets will release: published, and absent from the ignore array in .changeset/config.json.

    line 33
  2. Distinctive:

    If two bumps are both defensible, ask the user — the bump is a release decision, and once a release runs it is published.

    line 53
  3. Very specific:

    Create .changeset/<slug>.md.

    line 58
  4. Very specific:

    Commit the changeset alongside the change.

    line 89

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Write a changeset without the prompts

The interactive changeset add asks questions an agent has no way to answer — write the file yourself instead. The format is trivial; the judgment is in package selection and bump type.

1. Does this change need a changeset at all?

Docs, tooling, CI or test-only changes → an empty changeset: an explicit "nothing to release" that still satisfies the check. changeset add --empty is non-interactive and only writes the file, or write it yourself — an empty changeset is the frontmatter delimiters and nothing else:

---
---

Changes to published source → a real changeset, continue below.

2. Pick the packages

  • List the packages you directly changed and nothing else. Changesets bumps internal dependents on its own; listing them yourself over-bumps them and writes them a changelog entry describing work they didn't get.
  • Every name must be one Changesets will release: published, and absent from the ignore array in .changeset/config.json. Read that array — the directory tree doesn't reveal it.
  • Take each name from the package's own manifest. A directory name is not that name — the two diverge routinely, by a scope, a prefix, an abbreviation, or entirely — and a name matching no package fails the version step later, long after you've moved on.

3. Pick the bump

BumpWhen
majorBreaking change to a public API — a removed or renamed export, changed behavior consumers rely on
minorNew backwards-compatible behavior
patchBug fix or internal improvement with no API change

Decide the bump against the package's public surface, not against the description of the change. The surface is what consumers can reach — the declared entry points and the documented API — not everything the source exports. A changed signature, return shape, thrown error or default on that surface is breaking even though no export disappeared; a symbol reachable only by an internal path isn't on the surface at all. "Internal refactor" is a claim; the entry points are the evidence.

If two bumps are both defensible, ask the user — the bump is a release decision, and once a release runs it is published.

4. Write the file

Create .changeset/<slug>.md. Any unique kebab-case slug works; describe the change (fix-retry-timeout.md) so the file is readable in review.

---
"@scope/package-name": patch
---

Short imperative summary, written for whoever reads the changelog.

Optional body: what changed and why. For a breaking change, describe the migration a consumer has to make.

Quote the package names — an unquoted value starting with @ is not valid YAML. Multiple packages get one line each, and their bumps may differ.

The body is not read where you write it. It is copied verbatim into CHANGELOG.md, and into whatever the project publishes from there — release notes, a package page, a docs site. Every one of those renders it, so write a body that survives rendering.

One failure is invisible in review: <anything in angle brackets> reads as an HTML tag. A renderer that sanitizes deletes the tag and its contents; one that doesn't emits an element nothing displays. Either way the words are gone from the page while the source still reads correctly. Wrap placeholders in backticks — <why it matters> — and they render literally.

5. Verify

Changesets ships its own check: changeset status --since=<base-branch> prints what would be released and exits non-zero when a changed package has no changeset covering it. Run it however this project runs its installed tools, and confirm it reports nothing missing.

Commit the changeset alongside the change. changeset version applies the bumps and writes the changelog entries at release time — never hand-edit either.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.