Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate plugin dependency resolution logic with topological sorting.
.claude/skills/a5c-ai-plugin-dependency-resolver/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -55% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -42% | 0% |
Generate plugin dependency resolution logic.
typescriptinterface PluginNode { name: string; dependencies: string[]; } export function resolveDependencies(plugins: PluginNode[]): string[] { const graph = new Map<string, string[]>(); const inDegree = new Map<string, number>(); for (const plugin of plugins) { graph.set(plugin.name, plugin.dependencies); inDegree.set(plugin.name, 0); } for (const [, deps] of graph) { for (const dep of deps) { inDegree.set(dep, (inDegree.get(dep) || 0) + 1); } } const queue = [...inDegree.entries()].filter(([, d]) => d === 0).map(([n]) => n); const result: string[] = []; while (queue.length > 0) { const node = queue.shift()!; result.push(node); for (const dep of graph.get(node) || []) { inDegree.set(dep, inDegree.get(dep)! - 1); if (inDegree.get(dep) === 0) queue.push(dep); } } if (result.length !== plugins.length) { throw new Error('Circular dependency detected'); } return result.reverse(); }
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,657 | 9,986 | -14% | 1 | 1 | 0% | 2,414 | 2,689 | +11% | 0 | 0 | — |
case-02 | fail→fail | 12,991 | 12,734 | -2% | 1 | 1 | 0% | 2,824 | 2,995 | +6% | 0 | 0 | — |
case-03 | fail→pass | 17,144 | 12,522 | -27% | 1 | 1 | 0% | 3,284 | 3,128 | -5% | 0 | 0 | — |
case-04 | fail→fail | 12,018 | 9,378 | -22% | 1 | 1 | 0% | 2,378 | 2,206 | -7% | 0 | 0 | — |
case-05 | fail→fail | 9,937 | 6,040 | -39% | 1 | 1 | 0% | 2,147 | 1,644 | -23% | 0 | 0 | — |
case-06 | fail→fail | 5,545 | 4,858 | -12% | 1 | 1 | 0% | 1,136 | 1,332 | +17% | 0 | 0 | — |
case-07 | fail→pass | 10,267 | 2,979 | -71% | 1 | 1 | 0% | 1,824 | 809 | -56% | 0 | 0 | — |
case-08 | fail→fail | 12,723 | 9,299 | -27% | 1 | 1 | 0% | 2,971 | 2,185 | -26% | 0 | 0 | — |
case-09 | fail→pass | 9,092 | 1,997 | -78% | 1 | 1 | 0% | 1,486 | 666 | -55% | 0 | 0 | — |
case-10 | fail→pass | 14,830 | 7,519 | -49% | 1 | 1 | 0% | 2,471 | 1,494 | -40% | 0 | 0 | — |
case-11 | fail→pass | 10,881 | 4,358 | -60% | 1 | 1 | 0% | 2,090 | 1,204 | -42% | 0 | 0 | — |
case-12 | pass→pass | 9,581 | 3,578 | -63% | 1 | 1 | 0% | 1,874 | 980 | -48% | 0 | 0 | — |
case-13 | fail→fail | 8,452 | 5,794 | -31% | 1 | 1 | 0% | 1,559 | 1,433 | -8% | 0 | 0 | — |
case-14 | fail→pass | 9,387 | 4,775 | -49% | 1 | 1 | 0% | 1,785 | 1,288 | -28% | 0 | 0 | — |
case-15 | fail→fail | 4,829 | 2,933 | -39% | 1 | 1 | 0% | 796 | 843 | +6% | 0 | 0 | — |
case-16 | fail→pass | 9,789 | 11,291 | +15% | 1 | 1 | 0% | 1,756 | 2,154 | +23% | 0 | 0 | — |
case-17 | fail→pass | 9,440 | 2,701 | -71% | 1 | 1 | 0% | 1,751 | 620 | -65% | 0 | 0 | — |
case-18 | fail→fail | 10,855 | 7,422 | -32% | 1 | 1 | 0% | 1,852 | 1,853 | +0% | 0 | 0 | — |
case-19 | fail→fail | 11,112 | 8,237 | -26% | 1 | 1 | 0% | 2,005 | 1,880 | -6% | 0 | 0 | — |
case-20 | fail→fail | 14,572 | 13,653 | -6% | 1 | 1 | 0% | 3,514 | 3,233 | -8% | 0 | 0 | — |
case-21 | fail→fail | 10,217 | 5,533 | -46% | 1 | 1 | 0% | 2,022 | 1,496 | -26% | 0 | 0 | — |
case-22 | fail→fail | 15,367 | 10,093 | -34% | 1 | 1 | 0% | 3,256 | 2,804 | -14% | 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. The headline lift of +36 percentage points is the difference between those two pass rates over the 22 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.