Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Validate process JS files for correct SDK patterns, task definitions, syntax, and quality gate implementation.
.claude/skills/a5c-ai-process-validator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -21% | 0% |
You are process-validator - a specialized skill for validating Babysitter SDK process files for correct patterns and syntax.
This skill validates process JS files including:
javascript/** * @process specialization/process-name // Required * @description Process description // Required * @inputs { param: type } // Required * @outputs { result: type } // Required */
javascriptimport { defineTask } from '@a5c-ai/babysitter-sdk';
javascriptexport async function process(inputs, ctx) { // Destructure inputs const { param1, param2 = 'default' } = inputs; // Initialize artifacts const artifacts = []; // Use ctx.log for logging ctx.log('info', 'Starting process'); // Use ctx.task for task execution const result = await ctx.task(taskName, args); // Use ctx.breakpoint for approvals await ctx.breakpoint({ question, title, context }); // Return structured output return { success: true, artifacts }; }
javascriptexport const taskName = defineTask('task-name', (args, taskCtx) => ({ kind: 'agent', // Required: agent|skill|node|shell|breakpoint title: 'Task title', // Required: descriptive title skill: { name: 'skill-name' }, // Optional: skill reference agent: { // Required for kind: 'agent' name: 'agent-name', // Required: agent reference prompt: { // Required: prompt configuration role: 'Role', task: 'Task description', context: args, instructions: [], outputFormat: 'format' }, outputSchema: { // Required: JSON schema type: 'object', required: [], properties: {} } }, io: { // Required: io paths inputJsonPath: `tasks/${taskCtx.effectId}/input.json`, outputJsonPath: `tasks/${taskCtx.effectId}/result.json` }, labels: [] // Optional: categorization }));
| Rule | Description | |------|-------------| | HAS_JSDOC | File has JSDoc header | | HAS_IMPORT | Imports defineTask | | HAS_PROCESS | Exports process function | | VALID_TASKS | Task definitions are valid |
| Rule | Description | |------|-------------| | HAS_LOGGING | Uses ctx.log | | HAS_ARTIFACTS | Tracks artifacts | | HAS_RETURN | Returns structured output | | HAS_IO | Tasks have io configuration |
| Rule | Description | |------|-------------| | HAS_BREAKPOINTS | Has approval breakpoints | | HAS_QUALITY_GATES | Has quality scoring | | HAS_LABELS | Tasks have labels |
json{ "valid": true, "score": 95, "results": { "hasJsdoc": true, "hasImport": true, "hasProcessFunction": true, "taskCount": 5, "validTasks": 5, "hasLogging": true, "hasBreakpoints": true, "hasQualityGates": true }, "issues": [ { "severity": "warning", "rule": "HAS_LABELS", "message": "Task 'task-3' missing labels" } ], "artifacts": [] }
This skill integrates with:
process-creation.js - Post-generation validationspecialization-validator.js - Phase 3 validationphase3-implement-processes.js - Batch validation| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 23,043 | 19,064 | -17% | 1 | 1 | 0% | 3,653 | 2,259 | -38% | 0 | 0 | — |
case-02 | fail→pass | 20,076 | 11,197 | -44% | 1 | 1 | 0% | 3,088 | 3,322 | +8% | 0 | 0 | — |
case-03 | fail→fail | 18,729 | 8,318 | -56% | 1 | 1 | 0% | 2,942 | 2,670 | -9% | 0 | 0 | — |
case-04 | pass→pass | 15,269 | 18,098 | +19% | 1 | 1 | 0% | 2,423 | 5,314 | +119% | 0 | 0 | — |
case-05 | pass→fail | 20,579 | 15,922 | -23% | 1 | 1 | 0% | 3,008 | 3,940 | +31% | 0 | 0 | — |
case-06 | pass→pass | 10,319 | 10,206 | -1% | 1 | 1 | 0% | 2,053 | 2,962 | +44% | 0 | 0 | — |
case-07 | fail→pass | 10,062 | 4,570 | -55% | 1 | 1 | 0% | 1,462 | 1,926 | +32% | 0 | 0 | — |
case-08 | fail→pass | 7,547 | 5,373 | -29% | 1 | 1 | 0% | 1,276 | 1,805 | +41% | 0 | 0 | — |
case-09 | fail→pass | 13,716 | 4,939 | -64% | 1 | 1 | 0% | 2,376 | 1,879 | -21% | 0 | 0 | — |
case-10 | fail→pass | 12,378 | 7,157 | -42% | 1 | 1 | 0% | 2,303 | 2,099 | -9% | 0 | 0 | — |
case-11 | fail→pass | 19,852 | 5,823 | -71% | 1 | 1 | 0% | 3,392 | 2,196 | -35% | 0 | 0 | — |
case-12 | fail→pass | 12,510 | 5,824 | -53% | 1 | 1 | 0% | 1,829 | 1,935 | +6% | 0 | 0 | — |
case-13 | fail→fail | 18,919 | 7,332 | -61% | 1 | 1 | 0% | 2,527 | 2,375 | -6% | 0 | 0 | — |
case-14 | pass→pass | 11,309 | 6,466 | -43% | 1 | 1 | 0% | 1,674 | 2,020 | +21% | 0 | 0 | — |
case-15 | fail→pass | 7,694 | 4,875 | -37% | 1 | 1 | 0% | 1,162 | 1,793 | +54% | 0 | 0 | — |
case-16 | fail→pass | 13,085 | 8,093 | -38% | 1 | 1 | 0% | 1,744 | 2,312 | +33% | 0 | 0 | — |
case-17 | fail→fail | 9,743 | 3,582 | -63% | 1 | 1 | 0% | 1,498 | 1,690 | +13% | 0 | 0 | — |
case-18 | fail→pass | 7,879 | 4,256 | -46% | 1 | 1 | 0% | 1,058 | 1,910 | +81% | 0 | 0 | — |
case-19 | pass→pass | 14,904 | 6,692 | -55% | 1 | 1 | 0% | 2,081 | 2,081 | 0% | 0 | 0 | — |
case-20 | fail→pass | 14,757 | 9,497 | -36% | 1 | 1 | 0% | 2,355 | 2,748 | +17% | 0 | 0 | — |
case-21 | fail→pass | 17,317 | 3,926 | -77% | 1 | 1 | 0% | 2,570 | 1,864 | -27% | 0 | 0 | — |
case-22 | pass→pass | 11,027 | 4,380 | -60% | 1 | 1 | 0% | 1,938 | 1,906 | -2% | 0 | 0 | — |
case-23 | fail→pass | 11,451 | 6,653 | -42% | 1 | 1 | 0% | 1,854 | 1,908 | +3% | 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. 23 cases were attempted. The headline lift of +57 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
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.