Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Typeset professional documents with the Kami editorial design system only when
.claude/skills/cowork-os-kami/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 11% | 0% |
Typeset professional documents with the Kami editorial design system only when the user explicitly asks for Kami, the Kami skill, or the Kami editorial design system by name. Generic presentation, document, report, resume, PDF, and PowerPoint requests should use the native artifact tools instead.
This CoWork bundle ports the upstream Kami skill into a bundled skill with a workspace-local scaffold flow. The goal is to keep the strong visual system available on request while avoiding edits to bundled templates under resources/skills/.
TsangerJinKai02-W04.ttf.one-pagerlong-docletterportfolioresumeslidesdiagram-architecturediagram-flowchartdiagram-quadrantThe CoWork wrapper scaffolds a workspace-local project with a stable shape:
templates/ for document or slide source filesdiagrams/ for standalone diagram HTMLfonts/ for bundled open fonts and optional user-supplied licensed fontsoutputs/ for rendered PDF and PPTX filesmanifest.json for scaffold metadataEdit only the scaffolded project files. Do not edit bundled files under this skill directory.
scripts/setup.sh to detect local render support.scripts/bootstrap_project.py to scaffold a workspace-local project.CHEATSHEET.md or CHEATSHEET.en.md for light editsreferences/design*.md for new documents or larger layout workreferences/writing*.md for section structure and editorial tonereferences/production*.md for rendering issuesreferences/diagrams*.md for standalone diagram workscripts/render_html.py.scripts/render_slides.mjs, pointing it at the scaffolded slides.mjs or slides-en.mjs source file and the project's outputs/ directory.When upstream references mention TsangerJinKai02, treat that as an optional licensed upgrade, not a bundled requirement in this CoWork port. Default to the patched fallbacks already present in the scaffolded Chinese HTML templates unless the user explicitly provides a licensed font file.
CHEATSHEET.md / CHEATSHEET.en.md: quick style and template referencereferences/design.md / design.en.md: visual rules and layout tokensreferences/writing.md / writing.en.md: section structure and editorial guidancereferences/production.md / production.en.md: rendering and export troubleshootingreferences/diagrams.md / diagrams.en.md: standalone diagram guidancescripts/bootstrap_project.py: scaffold workspace-local source filesscripts/render_html.py: render a source HTML file to PDF with placeholder checksscripts/render_slides.mjs: render slide source to output.pptx and optionally output.pdfscripts/setup.sh: preflight local document-render dependenciesscripts/build.py: upstream bundle build/verification helper retained for reference../kami.json.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,233 | 7,248 | +0% | 1 | 1 | 0% | 1,089 | 1,382 | +27% | 0 | 0 | — |
case-02 | fail→fail | 29,190 | 7,973 | -73% | 1 | 1 | 0% | 6,172 | 1,978 | -68% | 0 | 0 | — |
case-03 | pass→pass | 23,299 | 27,579 | +18% | 1 | 1 | 0% | 4,624 | 6,843 | +48% | 0 | 0 | — |
case-04 | pass→pass | 12,477 | 8,985 | -28% | 1 | 1 | 0% | 1,959 | 2,554 | +30% | 0 | 0 | — |
case-05 | pass→pass | 18,158 | 24,438 | +35% | 1 | 1 | 0% | 3,086 | 5,703 | +85% | 0 | 0 | — |
case-06 | fail→pass | 12,408 | 6,360 | -49% | 1 | 1 | 0% | 1,992 | 2,021 | +1% | 0 | 0 | — |
case-07 | fail→pass | 12,715 | 4,555 | -64% | 1 | 1 | 0% | 2,044 | 1,751 | -14% | 0 | 0 | — |
case-08 | fail→pass | 11,533 | 6,201 | -46% | 1 | 1 | 0% | 1,983 | 2,073 | +5% | 0 | 0 | — |
case-09 | pass→pass | 4,487 | 3,208 | -29% | 1 | 1 | 0% | 698 | 1,491 | +114% | 0 | 0 | — |
case-10 | pass→pass | 11,949 | 5,645 | -53% | 1 | 1 | 0% | 1,869 | 1,996 | +7% | 0 | 0 | — |
case-11 | fail→pass | 7,260 | 2,376 | -67% | 1 | 1 | 0% | 1,089 | 1,404 | +29% | 0 | 0 | — |
case-12 | fail→pass | 7,738 | 2,402 | -69% | 1 | 1 | 0% | 1,280 | 1,418 | +11% | 0 | 0 | — |
case-13 | fail→pass | 5,317 | 3,150 | -41% | 1 | 1 | 0% | 850 | 1,620 | +91% | 0 | 0 | — |
case-14 | fail→pass | 11,531 | 1,787 | -85% | 1 | 1 | 0% | 2,034 | 1,296 | -36% | 0 | 0 | — |
case-15 | fail→pass | 4,897 | 3,342 | -32% | 1 | 1 | 0% | 765 | 1,649 | +116% | 0 | 0 | — |
case-16 | fail→pass | 14,460 | 2,921 | -80% | 1 | 1 | 0% | 2,240 | 1,421 | -37% | 0 | 0 | — |
case-17 | fail→pass | 12,506 | 3,257 | -74% | 1 | 1 | 0% | 2,000 | 1,454 | -27% | 0 | 0 | — |
case-18 | fail→pass | 11,906 | 6,629 | -44% | 1 | 1 | 0% | 1,967 | 1,984 | +1% | 0 | 0 | — |
case-19 | pass→pass | 11,281 | 7,668 | -32% | 1 | 1 | 0% | 1,833 | 2,175 | +19% | 0 | 0 | — |
case-20 | fail→pass | 7,573 | 2,403 | -68% | 1 | 1 | 0% | 1,163 | 1,408 | +21% | 0 | 0 | — |
case-21 | pass→pass | 7,520 | 1,757 | -77% | 1 | 1 | 0% | 1,111 | 1,307 | +18% | 0 | 0 | — |
case-22 | pass→pass | 7,639 | 2,680 | -65% | 1 | 1 | 0% | 1,093 | 1,427 | +31% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.