Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when analyzing a reference PPTX for read-only structure, theme, typography, layout rhythm, diagnostics, derived template catalogs, or safe OOXML package inspection.
.claude/skills/wshobson-pptx-reference-deck-analysis/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-23 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -73% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 49% | 0% |
Inspect a reference deck as design evidence. This skill never copies, clones, or mutates a source deck.
Implement required extraction on demand with a small task-local python-pptx script. Capture only the information required for the new deck:
summary, slides, and read-only layout_tree evidence;Install defusedxml from requirements.txt before using the bundled utilities.
scripts/inspect.py <deck.pptx> for a compact JSON report of slide order, text, theme tokens, relationships, notes, comments, animations, and media.scripts/validate_package.py <deck.pptx> --output <report.json> for malformed XML, broken internal relationships, content-type gaps, duplicate layout links, and orphaned parts.scripts/unpack.py <deck.pptx> <output-dir> only when raw-package evidence is necessary..rels owner; never infer slide order from filenames.defusedxml; do not enable entity expansion, DTD loading, or network access.See references/reference-deck-analysis.md for output shapes, references/reference-deck-analysis-patterns.md for documentation-only patterns, and references/ooxml-parsing.md for package part maps.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-23 | fail→pass | 8,599 | 6,428 | -25% | 1 | 1 | 0% | 1,461 | 1,887 | +29% | 0 | 0 | — |
case-01 | fail→fail | 10,331 | 5,670 | -45% | 1 | 1 | 0% | 727 | 944 | +30% | 0 | 0 | — |
case-02 | fail→fail | 27,189 | 6,067 | -78% | 1 | 1 | 0% | 3,994 | 975 | -76% | 0 | 0 | — |
case-03 | fail→fail | 16,579 | 6,830 | -59% | 1 | 1 | 0% | 3,109 | 973 | -69% | 0 | 0 | — |
case-04 | fail→pass | 27,837 | 5,697 | -80% | 1 | 1 | 0% | 6,190 | 1,672 | -73% | 0 | 0 | — |
case-05 | fail→fail | 6,859 | 8,647 | +26% | 1 | 1 | 0% | 250 | 1,333 | +433% | 0 | 0 | — |
case-06 | pass→pass | 7,642 | 7,573 | -1% | 1 | 1 | 0% | 1,820 | 2,274 | +25% | 0 | 0 | — |
case-07 | fail→pass | 11,417 | 2,541 | -78% | 1 | 1 | 0% | 2,270 | 1,077 | -53% | 0 | 0 | — |
case-08 | fail→pass | 10,085 | 6,822 | -32% | 1 | 1 | 0% | 2,131 | 1,990 | -7% | 0 | 0 | — |
case-09 | fail→pass | 4,077 | 2,607 | -36% | 1 | 1 | 0% | 772 | 1,147 | +49% | 0 | 0 | — |
case-10 | pass→pass | 9,427 | 4,037 | -57% | 1 | 1 | 0% | 1,829 | 1,328 | -27% | 0 | 0 | — |
case-11 | fail→pass | 8,929 | 3,739 | -58% | 1 | 1 | 0% | 1,600 | 1,159 | -28% | 0 | 0 | — |
case-12 | fail→pass | 6,075 | 1,418 | -77% | 1 | 1 | 0% | 1,150 | 805 | -30% | 0 | 0 | — |
case-13 | pass→pass | 13,772 | 5,176 | -62% | 1 | 1 | 0% | 2,431 | 1,474 | -39% | 0 | 0 | — |
case-14 | fail→pass | 2,136 | 941 | -56% | 1 | 1 | 0% | 359 | 713 | +99% | 0 | 0 | — |
case-15 | fail→pass | 11,974 | 8,737 | -27% | 1 | 1 | 0% | 2,517 | 2,448 | -3% | 0 | 0 | — |
case-16 | pass→pass | 9,444 | 2,332 | -75% | 1 | 1 | 0% | 1,747 | 967 | -45% | 0 | 0 | — |
case-17 | fail→pass | 14,000 | 11,231 | -20% | 1 | 1 | 0% | 2,360 | 2,249 | -5% | 0 | 0 | — |
case-18 | fail→pass | 10,695 | 9,876 | -8% | 1 | 1 | 0% | 2,115 | 2,554 | +21% | 0 | 0 | — |
case-19 | pass→pass | 18,036 | 17,849 | -1% | 1 | 1 | 0% | 2,949 | 3,459 | +17% | 0 | 0 | — |
case-20 | pass→pass | 11,010 | 7,690 | -30% | 1 | 1 | 0% | 2,188 | 2,209 | +1% | 0 | 0 | — |
case-21 | pass→pass | 5,639 | 1,825 | -68% | 1 | 1 | 0% | 907 | 873 | -4% | 0 | 0 | — |
case-22 | pass→pass | 73,186 | 3,190 | -96% | 1 | 1 | 0% | 2,248 | 1,127 | -50% | 0 | 0 | — |
case-24 | pass→pass | 11,702 | 7,491 | -36% | 1 | 1 | 0% | 2,552 | 2,158 | -15% | 0 | 0 | — |
case-25 | fail→pass | 15,620 | 2,324 | -85% | 1 | 1 | 0% | 3,180 | 991 | -69% | 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. 25 cases were attempted, and 21 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +48 percentage points is the difference between those two pass rates over the 21 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.