Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate polished dark-themed architecture diagrams as standalone HTML files with inline SVG, semantic system grouping, and readable connection flows.
.claude/skills/cowork-os-architecture-diagram/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -36% | 0% |
Turn an architecture description or inspected codebase into a self-contained HTML diagram with inline SVG that opens directly in any modern browser.
| Name | Type | Required | Description | |---|---|---|---| | system_name | string | No | Short system or project name for the header | | diagram_brief | string | No | Structured architecture description | | diagram_type | select | No | Diagram emphasis: auto, system-overview, cloud-infrastructure, deployment, data-flow, security, network-topology, microservices | | output_path | string | No | Workspace-relative or absolute HTML output path |
assets/template.html. Do not freehand the outer page shell unless the request explicitly needs a very different composition.references/full-guidance.md before major layout work. It contains the spacing, legend placement, component taxonomy, and arrow-order rules that keep the diagrams readable.viewBox is cheap; unreadable overlaps are not.::html{...} directive that points at the generated HTML file so the main conversation view can render the diagram inline.references/full-guidance.md: design system, layout heuristics, and output checklistassets/template.html: starter HTML/SVG template to customizeLICENSE.txt: upstream MIT license carried with the bundled port../architecture-diagram.json.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-21 | pass→pass | 11,611 | 8,071 | -30% | 1 | 1 | 0% | 1,853 | 1,858 | +0% | 0 | 0 | — |
case-11 | pass→pass | 11,524 | 3,970 | -66% | 1 | 1 | 0% | 1,253 | 1,029 | -18% | 0 | 0 | — |
case-01 | fail→fail | 27,468 | 29,765 | +8% | 1 | 1 | 0% | 6,218 | 6,871 | +11% | 0 | 0 | — |
case-02 | fail→fail | 31,650 | 3,044 | -90% | 1 | 1 | 0% | 6,187 | 826 | -87% | 0 | 0 | — |
case-03 | fail→fail | 31,240 | 29,494 | -6% | 1 | 1 | 0% | 6,193 | 6,847 | +11% | 0 | 0 | — |
case-04 | fail→pass | 13,249 | 4,044 | -69% | 1 | 1 | 0% | 1,974 | 1,157 | -41% | 0 | 0 | — |
case-05 | fail→pass | 8,518 | 2,856 | -66% | 1 | 1 | 0% | 1,293 | 1,082 | -16% | 0 | 0 | — |
case-06 | fail→pass | 39,523 | 30,161 | -24% | 1 | 1 | 0% | 2,367 | 2,926 | +24% | 0 | 0 | — |
case-20 | pass→pass | 9,731 | 8,057 | -17% | 1 | 1 | 0% | 1,651 | 2,085 | +26% | 0 | 0 | — |
case-07 | pass→pass | 11,808 | 5,958 | -50% | 1 | 1 | 0% | 1,836 | 1,513 | -18% | 0 | 0 | — |
case-08 | pass→pass | 8,691 | 9,119 | +5% | 1 | 1 | 0% | 1,272 | 1,536 | +21% | 0 | 0 | — |
case-09 | fail→pass | 14,676 | 1,988 | -86% | 1 | 1 | 0% | 1,321 | 933 | -29% | 0 | 0 | — |
case-10 | fail→pass | 26,945 | 2,650 | -90% | 1 | 1 | 0% | 1,635 | 1,040 | -36% | 0 | 0 | — |
case-12 | fail→pass | 8,325 | 1,799 | -78% | 1 | 1 | 0% | 1,439 | 872 | -39% | 0 | 0 | — |
case-13 | fail→pass | 12,865 | 4,467 | -65% | 1 | 1 | 0% | 1,823 | 1,395 | -23% | 0 | 0 | — |
case-14 | pass→pass | 7,954 | 3,978 | -50% | 1 | 1 | 0% | 1,182 | 1,284 | +9% | 0 | 0 | — |
case-15 | pass→pass | 13,954 | 3,551 | -75% | 1 | 1 | 0% | 1,612 | 1,127 | -30% | 0 | 0 | — |
case-16 | fail→pass | 10,156 | 2,860 | -72% | 1 | 1 | 0% | 1,566 | 1,077 | -31% | 0 | 0 | — |
case-17 | fail→pass | 9,786 | 5,753 | -41% | 1 | 1 | 0% | 1,452 | 1,459 | +0% | 0 | 0 | — |
case-18 | fail→pass | 5,026 | 2,779 | -45% | 1 | 1 | 0% | 832 | 1,006 | +21% | 0 | 0 | — |
case-19 | pass→pass | 13,387 | 5,163 | -61% | 1 | 1 | 0% | 2,053 | 1,340 | -35% | 0 | 0 | — |
case-22 | pass→pass | 27,820 | 26,208 | -6% | 1 | 1 | 0% | 6,163 | 6,817 | +11% | 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 +45 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.