[rqbazan]

Summary

The skill's description told agents to use it "before designing anything public". That fired for any public work, so agents loaded Variable Lab's brand for work outside Variable Lab, Variable Land's own products included. The description and the skill's opening now scope it to Variable Lab and its Projects.

skills/vlab-brand/SKILL.md, the description:

 Variable Lab's brand guideline - principles, Logo, color, type, voice, UI icons and how a Project's Sign is drawn.
-Use before designing anything public, writing public copy (a README, a page, an image) or giving a Project its Sign.
+Use only for Variable Lab and its open source Projects, before designing anything public, writing public copy (a README, a page, an image) or giving a Project its Sign.
+Variable Land's own products use their own brand.

The opening paragraph:

 Variable Lab is Variable Land's open source arm.
-This guideline holds its rules for anything public: sites, READMEs, package pages, images and icons.
-It states rules, never one Project's specifics: ...
+This guideline is for Variable Lab and its Projects only, and holds their rules for anything public: sites, READMEs, package pages, images and icons.
+Variable Land's own products use their own brand.
+The guideline states rules, never one Project's specifics: ...

No other chapter changes. The changeset (.changeset/scope-to-variable-lab.md, a patch) is there because the contributing guide asks for one on any change to skills/.

Evidence

The description, read the way an agent reads it when choosing a skill:

Task	Before	After
A Variable Lab Project's README or Sign	fires	fires
Variable Lab's own site, or this repository	fires	fires
A page for one of Variable Land's own products	fires	doesn't fire: "Variable Land's own products use their own brand"
A README in a repository of neither	fires	doesn't fire: "only for Variable Lab and its open source Projects"

• pnpm check passes. It doesn't read SKILL.md, so the frontmatter was also parsed as strict YAML (yaml 2.9.1). The description holds no : , which would break a plain scalar.
• Both commits are signed and verified.

Merge Danger

Door: two-way

A text change. Reverting it restores the old description.

Blast Radius: triggering

Agents install the skill from main, so the new description reaches them on their next install or update, before any release. They load it for Variable Lab work only. The next "chore: update versions" pull request publishes it as a patch.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EGJDrvjdi8RDE89MahhuaJ

[pullfrog]

New pull request. Leaping into action...

  ^{｜ View workflow run ｜ via Pullfrog ｜ Using grok-4.7 ｜ 𝕏}

[changeset-bot]

🦋 Changeset detected

Latest commit: 25ee9c0

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@variablelab/brand	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR