token-normalizer — Claude Code skill: token proposals and setup
token-normalizer is a Claude Code skill that turns repeated implementation values into reviewable semantic token proposals. It keeps existing names and flags exceptions instead of forcing merges.
MIT ★ 529 updated 2026-11-17
npx skills add ShuaiMXu/happyhands --skill token-normalizer -a claude-codeWhat token-normalizer does
token-normalizer collects repeated implementation values into a semantic token proposal with supporting evidence. It needs raw values, existing token aliases, and the locations where those values are used.
The skill preserves current semantic names, lists each proposed merge, and flags exceptions rather than forcing them together. Frequency by itself does not establish that two values mean the same thing.
Its output includes structured evidence, findings, proposed updates, and limitations. If required inputs or professional review are missing, it uses needs-evidence; it does not turn observations from one project into universal rules.
What it helps with
- group repeated visual values into token proposals
- compare raw values with existing token aliases
- list proposed merges for review
- flag exceptions without forcing a merge
Example prompts for the agent
- Map these repeated visual values to semantic tokens and show the evidence for each proposal.
- Compare these values with existing token aliases and flag possible merges.
- Draft a token proposal with usage locations and list the exceptions.
How to install token-normalizer
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill token-normalizer -a claude-code2Manually: put the files in place
~/.claude/skills/token-normalizer/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/token-normalizer/
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill token-normalizer -a codex2Manually: put the files in place
.agents/skills/token-normalizer/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill token-normalizer -a cursor2Manually: put the files in place
.cursor/skills/token-normalizer/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ShuaiMXu/happyhands skills/product-design/token-normalizer2Manually: put the files in place
.github/skills/token-normalizer/SKILL.md1Skills CLI (Vercel)
npx skills add ShuaiMXu/happyhands --skill token-normalizer -a gemini-cli2Platform CLI
gemini skills install https://github.com/ShuaiMXu/happyhands --path skills/product-design/token-normalizerManually: put the files in place
.devin/skills/token-normalizer/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about token-normalizer
How do I install token-normalizer in Claude Code?
Run npx skills add ShuaiMXu/happyhands --skill token-normalizer -a claude-code. The skill file is at ~/.claude/skills/token-normalizer/SKILL.md.
What does token-normalizer do?
It turns repeated implementation values into semantic token proposals with evidence for review. It lists proposed merges and flags exceptions.
What inputs does token-normalizer need?
It needs raw visual values, existing token aliases, and usage locations. When required inputs or professional review are missing, it uses needs-evidence.
Source text
SKILL.md · MIT
--- name: token-normalizer description: "Map raw visual values and aliases into candidate semantic design tokens without silently changing the product." --- # Token Normalizer Turn repeated implementation values into a reviewable semantic token proposal. ## Required evidence - raw values - existing token aliases - usage locations ## Decisions - Preserve existing semantic names - Show every proposed merge - Flag exceptions instead of forcing them ## Boundary Frequency alone does not determine semantic equivalence. Return structured evidence, findings, proposed updates and limitations. Use `needs-evidence` when required inputs or professional review are missing. Never invent an Experience Score or promote a project observation into a universal rule.
© ShuaiMXu · MIT · ShuaiMXu/happyhands