Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →PDF accessibility remediator. Generates scripts for programmatic fixes via pdf-lib/qpdf/ghostscript (title, language, reading order, tags, alt text) and provides step-by-step Adobe Acrobat Pro instructions for manual fixes (table structure, complex layouts, form tooltips).
.claude/skills/community-access-pdf-remediator-363484/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 1% | 0% |
You fix accessibility issues in PDF documents, separating fixes into programmatic (script-based) and manual (Acrobat Pro) categories.
| Issue | Tool | |-------|------| | Missing document title | pdf-lib | | Missing document language | qpdf | | Missing reading order | qpdf | | Incorrect tag types | qpdf | | Decorative images not artifact | qpdf | | Missing alt text on figures | pdf-lib | | Missing PDF/UA identifier | pdf-lib | | Missing XMP metadata | pdf-lib |
| Issue | Tool Required | |-------|---------------| | Table structure | Acrobat Pro Tags panel | | Form field tooltips | Acrobat Pro Forms editor | | Complex reading order | Acrobat Pro Order panel | | Abbreviation text | Acrobat Pro Tags panel | | Contrast in images | Image editor + re-embed | | Bookmark structure | Acrobat Pro Bookmarks panel |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→pass | 20,126 | 10,212 | -49% | 1 | 1 | 0% | 3,274 | 1,992 | -39% | 0 | 0 | — |
case-08 | pass→pass | 34,504 | 8,541 | -75% | 1 | 1 | 0% | 2,568 | 1,496 | -42% | 0 | 0 | — |
case-09 | fail→pass | 20,902 | 13,937 | -33% | 1 | 1 | 0% | 3,043 | 2,383 | -22% | 0 | 0 | — |
case-06 | pass→pass | 17,113 | 9,878 | -42% | 1 | 1 | 0% | 2,443 | 1,751 | -28% | 0 | 0 | — |
case-01 | fail→pass | 18,064 | 10,701 | -41% | 1 | 1 | 0% | 3,014 | 2,087 | -31% | 0 | 0 | — |
case-02 | fail→pass | 16,298 | 12,545 | -23% | 1 | 1 | 0% | 2,894 | 2,548 | -12% | 0 | 0 | — |
case-03 | fail→fail | 12,722 | 11,935 | -6% | 1 | 1 | 0% | 2,101 | 2,487 | +18% | 0 | 0 | — |
case-04 | fail→fail | 12,721 | 10,064 | -21% | 1 | 1 | 0% | 2,115 | 2,046 | -3% | 0 | 0 | — |
case-05 | fail→pass | 17,085 | 13,361 | -22% | 1 | 1 | 0% | 3,000 | 3,035 | +1% | 0 | 0 | — |
case-10 | pass→pass | 14,029 | 7,333 | -48% | 1 | 1 | 0% | 1,972 | 1,427 | -28% | 0 | 0 | — |
case-11 | fail→pass | 19,491 | 19,383 | -1% | 1 | 1 | 0% | 3,602 | 3,566 | -1% | 0 | 0 | — |
case-12 | fail→pass | 14,659 | 7,875 | -46% | 1 | 1 | 0% | 2,091 | 1,590 | -24% | 0 | 0 | — |
case-13 | fail→pass | 8,718 | 7,471 | -14% | 1 | 1 | 0% | 1,372 | 1,483 | +8% | 0 | 0 | — |
case-14 | fail→pass | 14,780 | 12,767 | -14% | 1 | 1 | 0% | 2,173 | 2,372 | +9% | 0 | 0 | — |
case-15 | fail→pass | 12,330 | 7,906 | -36% | 1 | 1 | 0% | 2,143 | 1,684 | -21% | 0 | 0 | — |
case-16 | fail→fail | 8,539 | 5,297 | -38% | 1 | 1 | 0% | 1,390 | 1,112 | -20% | 0 | 0 | — |
case-17 | fail→pass | 9,278 | 6,627 | -29% | 1 | 1 | 0% | 1,694 | 1,382 | -18% | 0 | 0 | — |
case-18 | fail→pass | 10,774 | 7,576 | -30% | 1 | 1 | 0% | 1,859 | 1,672 | -10% | 0 | 0 | — |
case-19 | fail→pass | 8,446 | 2,399 | -72% | 1 | 1 | 0% | 1,347 | 633 | -53% | 0 | 0 | — |
case-20 | pass→pass | 12,535 | 10,514 | -16% | 1 | 1 | 0% | 2,738 | 2,643 | -3% | 0 | 0 | — |
case-21 | pass→pass | 9,099 | 4,919 | -46% | 1 | 1 | 0% | 1,504 | 1,050 | -30% | 0 | 0 | — |
case-22 | pass→pass | 14,397 | 10,168 | -29% | 1 | 1 | 0% | 2,527 | 2,093 | -17% | 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 +59 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.