Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert at handling file uploads and cloud storage. Covers S3, Cloudflare R2, presigned URLs, multipart uploads, and image optimization. Knows how to handle large files without blocking. Use when: file upload, S3, R2, presigned URL, multipart.
.claude/skills/davila7-file-uploads/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -5% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | -17% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 23% | 0% |
Role: File Upload Specialist
Careful about security and performance. Never trusts file extensions. Knows that large uploads need special handling. Prefers presigned URLs over server proxying.
| Issue | Severity | Solution | |-------|----------|----------| | Trusting client-provided file type | critical | # CHECK MAGIC BYTES | | No upload size restrictions | high | # SET SIZE LIMITS | | User-controlled filename allows path traversal | critical | # SANITIZE FILENAMES | | Presigned URL shared or cached incorrectly | medium | # CONTROL PRESIGNED URL DISTRIBUTION |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,859 | 18,048 | +1% | 1 | 1 | 0% | 3,251 | 3,291 | +1% | 0 | 0 | — |
case-02 | fail→pass | 23,275 | 21,931 | -6% | 1 | 1 | 0% | 3,867 | 3,791 | -2% | 0 | 0 | — |
case-03 | fail→pass | 25,082 | 21,132 | -16% | 1 | 1 | 0% | 4,827 | 4,247 | -12% | 0 | 0 | — |
case-04 | pass→pass | 17,589 | 15,357 | -13% | 1 | 1 | 0% | 3,035 | 2,898 | -5% | 0 | 0 | — |
case-05 | pass→pass | 15,007 | 10,978 | -27% | 1 | 1 | 0% | 2,557 | 2,112 | -17% | 0 | 0 | — |
case-06 | pass→pass | 12,823 | 14,205 | +11% | 1 | 1 | 0% | 2,133 | 2,627 | +23% | 0 | 0 | — |
case-07 | pass→pass | 16,580 | 15,351 | -7% | 1 | 1 | 0% | 2,604 | 2,716 | +4% | 0 | 0 | — |
case-08 | pass→pass | 15,817 | 10,854 | -31% | 1 | 1 | 0% | 2,415 | 1,862 | -23% | 0 | 0 | — |
case-09 | pass→pass | 18,825 | 16,170 | -14% | 1 | 1 | 0% | 3,112 | 3,037 | -2% | 0 | 0 | — |
case-10 | pass→pass | 17,321 | 29,362 | +70% | 1 | 1 | 0% | 3,129 | 3,680 | +18% | 0 | 0 | — |
case-11 | pass→pass | 10,482 | 8,016 | -24% | 1 | 1 | 0% | 1,574 | 1,436 | -9% | 0 | 0 | — |
case-12 | pass→pass | 16,158 | 13,168 | -19% | 1 | 1 | 0% | 2,601 | 2,472 | -5% | 0 | 0 | — |
case-13 | pass→pass | 8,463 | 6,881 | -19% | 1 | 1 | 0% | 1,355 | 1,150 | -15% | 0 | 0 | — |
case-14 | pass→pass | 13,270 | 10,617 | -20% | 1 | 1 | 0% | 2,148 | 1,823 | -15% | 0 | 0 | — |
case-15 | pass→pass | 15,110 | 9,756 | -35% | 1 | 1 | 0% | 2,312 | 1,814 | -22% | 0 | 0 | — |
case-16 | pass→pass | 12,660 | 15,451 | +22% | 1 | 1 | 0% | 2,432 | 2,742 | +13% | 0 | 0 | — |
case-17 | pass→pass | 15,320 | 18,846 | +23% | 1 | 1 | 0% | 2,654 | 2,754 | +4% | 0 | 0 | — |
case-18 | pass→pass | 8,814 | 7,268 | -18% | 1 | 1 | 0% | 1,353 | 1,127 | -17% | 0 | 0 | — |
case-19 | pass→pass | 13,841 | 13,492 | -3% | 1 | 1 | 0% | 2,203 | 2,258 | +2% | 0 | 0 | — |
case-20 | pass→pass | 6,309 | 7,333 | +16% | 1 | 1 | 0% | 1,277 | 1,622 | +27% | 0 | 0 | — |
case-21 | pass→pass | 5,667 | 6,369 | +12% | 1 | 1 | 0% | 1,136 | 1,211 | +7% | 0 | 0 | — |
case-22 | pass→pass | 8,264 | 7,026 | -15% | 1 | 1 | 0% | 1,580 | 1,478 | -6% | 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 +9 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.