Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when authoring or repairing a coordinate-explicit JSON specification for an editable PPTX deck.
.claude/skills/wshobson-pptx-slide-specification/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 8% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -20% | 0% |
Author final coordinates directly in layout_tree. No renderer may decide placement, shrink text, or infer layout after the audit.
summary and slides.id, a message-led title, an accessibility reading order, and a complete layout_tree.id, kind, role, classification, content, style, bbox, and z_index.summary contains an explicit layout_policy (safe margin, content bottom, footer top, minimum gap) and accessibility metadata.layout_design objects may be full bleed.text, shape, line, table, and image objects. Recreate labels and values from any supporting visual as native objects.source_ref for sourced claims.A task-local builder starts from a blank slide layout and maps all bboxes with Inches(...). Explicitly set wrapping, disabled auto-size, text insets, anchors, alignment, fonts, colors, line settings, image aspect ratio, and hidden-slide state. Reject zero or negative geometry before adding an object.
See references/layout-contract.md for the compact schema and repair guidance.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 24,438 | 25,621 | +5% | 1 | 1 | 0% | 6,203 | 6,624 | +7% | 0 | 0 | — |
case-12 | fail→pass | 9,259 | 7,521 | -19% | 1 | 1 | 0% | 1,844 | 2,068 | +12% | 0 | 0 | — |
case-02 | fail→pass | 21,649 | 82,614 | +282% | 1 | 1 | 0% | 5,627 | 6,624 | +18% | 0 | 0 | — |
case-03 | fail→fail | 22,499 | 23,003 | +2% | 1 | 1 | 0% | 6,194 | 6,616 | +7% | 0 | 0 | — |
case-04 | pass→pass | 26,075 | 23,051 | -12% | 1 | 1 | 0% | 6,197 | 6,619 | +7% | 0 | 0 | — |
case-05 | fail→fail | 13,301 | 12,054 | -9% | 1 | 1 | 0% | 3,050 | 3,493 | +15% | 0 | 0 | — |
case-06 | fail→fail | 6,132 | 21,924 | +258% | 1 | 1 | 0% | 1,171 | 6,603 | +464% | 0 | 0 | — |
case-07 | fail→pass | 19,466 | 19,207 | -1% | 1 | 1 | 0% | 4,008 | 4,310 | +8% | 0 | 0 | — |
case-08 | fail→fail | 11,119 | 11,341 | +2% | 1 | 1 | 0% | 2,446 | 2,969 | +21% | 0 | 0 | — |
case-09 | pass→pass | 20,561 | 24,244 | +18% | 1 | 1 | 0% | 5,136 | 6,607 | +29% | 0 | 0 | — |
case-10 | fail→pass | 13,885 | 8,598 | -38% | 1 | 1 | 0% | 2,690 | 2,151 | -20% | 0 | 0 | — |
case-11 | fail→pass | 10,005 | 2,449 | -76% | 1 | 1 | 0% | 1,659 | 826 | -50% | 0 | 0 | — |
case-13 | pass→pass | 7,408 | 4,773 | -36% | 1 | 1 | 0% | 1,262 | 1,264 | +0% | 0 | 0 | — |
case-14 | pass→pass | 12,705 | 9,583 | -25% | 1 | 1 | 0% | 2,715 | 2,129 | -22% | 0 | 0 | — |
case-15 | fail→pass | 13,223 | 8,453 | -36% | 1 | 1 | 0% | 2,454 | 2,049 | -17% | 0 | 0 | — |
case-16 | pass→pass | 17,958 | 13,508 | -25% | 1 | 1 | 0% | 3,797 | 3,380 | -11% | 0 | 0 | — |
case-17 | pass→pass | 11,261 | 25,541 | +127% | 1 | 1 | 0% | 2,498 | 2,524 | +1% | 0 | 0 | — |
case-18 | pass→pass | 9,290 | 8,022 | -14% | 1 | 1 | 0% | 1,945 | 2,300 | +18% | 0 | 0 | — |
case-19 | pass→fail | 6,620 | 23,955 | +262% | 1 | 1 | 0% | 1,328 | 6,603 | +397% | 0 | 0 | — |
case-20 | pass→pass | 9,269 | 9,329 | +1% | 1 | 1 | 0% | 2,209 | 2,640 | +20% | 0 | 0 | — |
case-21 | pass→fail | 10,957 | 7,658 | -30% | 1 | 1 | 0% | 2,021 | 1,831 | -9% | 0 | 0 | — |
case-22 | pass→pass | 13,992 | 12,272 | -12% | 1 | 1 | 0% | 3,203 | 2,921 | -9% | 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 +23 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are 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.