skills/uploadthing/_scores.json

{
  "version": "1.0.0",
  "skillHash": "sha256:3a3e645202c7e04cefea61834e5ef290b653d305f6a0b071f730b7c9e910081e",
  "scoredAt": "2026-05-13T10:32:08.243Z",
  "backend": "ollama",
  "model": "gpt-oss:20b",
  "quality": {
    "score": 93,
    "dimensions": {
      "clarity": "PASS",
      "completeness": "WEAK",
      "conciseness": "PASS",
      "actionability": "PASS",
      "crossPlatform": "PASS",
      "examples": "PASS"
    },
    "issues": [
      {
        "severity": "MEDIUM",
        "category": "completeness",
        "detail": "The skill does not cover error scenarios and edge cases beyond authentication."
      }
    ]
  },
  "security": {
    "verdict": "SAFE",
    "issues": []
  },
  "impact": {
    "baselineAvg": 35,
    "treatmentAvg": 89,
    "multiplier": 2.54,
    "scenarios": [
      {
        "name": "Authenticated avatar upload writes user.avatarUrl",
        "baseline": 35,
        "treatment": 90,
        "rationale": "Treatment defines a FileRouter with createUploadthing(), f({ image: { maxFileSize: '2MB', maxFileCount: 1 } }), session-checking .middleware, and .onUploadComplete that calls db.user.update, then renders UploadButton<OurFileRouter, 'avatarUploader'>; baseline proposes manual S3 presigned URLs and misses the FileRouter DSL."
      },
      {
        "name": "Drag-and-drop PDF dropzone with 5x10MB limit",
        "baseline": 35,
        "treatment": 88,
        "rationale": "Treatment uses f({ pdf: { maxFileSize: '10MB', maxFileCount: 5 } }) with a session-guarded .middleware and .onUploadComplete persisting via db.document.create, plus UploadDropzone with the typed endpoint; baseline reaches for react-dropzone + raw S3 multipart and never wires UploadThing's typed components."
      }
    ],
    "backend": "subagent",
    "model": "claude-sonnet-4-6"
  }
}