Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build, update, and apply iOS design specifications using Apple Human Interface Guidelines (HIG) source data. Use when a task asks for iOS UI/UX rules, Apple design standards, component behavior, accessibility constraints, interaction patterns, or feature-level design-spec writing grounded in official HIG pages.
.claude/skills/davepoon-ios-hig-design-guide/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -6% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -8% | 0% |
Use this skill to produce iOS design recommendations that stay close to official Apple guidance.
Run:
bashpython3 scripts/sync_apple_hig_sources.py --skill-dir .
references/apple-hig-ios-raw.mdreferences/apple-hig-ios-fulltext.mdreferences/apple-hig-ios-curated.mdreferences/ios-design-spec-workflow.mdreferences/raw/pages/design/human-interface-guidelines/*.jsonreferences/raw/catalog.jsondownload_error is 0 in references/raw/catalog.json./design/human-interface-guidelines/designing-for-ios.references/apple-hig-ios-curated.md for day-to-day use; use full dump only when needed.For each selected page, pull concrete rules into implementable statements:
Default output structure:
references/; do not hand-edit generated outputs.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,952 | 22,995 | +5% | 1 | 1 | 0% | 3,714 | 4,616 | +24% | 0 | 0 | — |
case-02 | fail→fail | 33,750 | 22,670 | -33% | 1 | 1 | 0% | 6,193 | 4,393 | -29% | 0 | 0 | — |
case-03 | fail→pass | 23,362 | 19,102 | -18% | 1 | 1 | 0% | 3,944 | 3,885 | -1% | 0 | 0 | — |
case-04 | pass→pass | 20,037 | 22,019 | +10% | 1 | 1 | 0% | 3,749 | 4,620 | +23% | 0 | 0 | — |
case-05 | pass→pass | 15,178 | 23,880 | +57% | 1 | 1 | 0% | 3,189 | 5,752 | +80% | 0 | 0 | — |
case-06 | pass→pass | 25,044 | 24,280 | -3% | 1 | 1 | 0% | 4,251 | 4,734 | +11% | 0 | 0 | — |
case-07 | fail→pass | 24,869 | 20,941 | -16% | 1 | 1 | 0% | 4,620 | 4,324 | -6% | 0 | 0 | — |
case-08 | fail→pass | 19,615 | 24,177 | +23% | 1 | 1 | 0% | 3,431 | 4,231 | +23% | 0 | 0 | — |
case-09 | fail→fail | 18,695 | 18,706 | +0% | 1 | 1 | 0% | 3,348 | 3,242 | -3% | 0 | 0 | — |
case-10 | fail→fail | 14,870 | 18,877 | +27% | 1 | 1 | 0% | 3,046 | 3,731 | +22% | 0 | 0 | — |
case-11 | fail→pass | 23,775 | 17,450 | -27% | 1 | 1 | 0% | 3,984 | 3,648 | -8% | 0 | 0 | — |
case-12 | fail→pass | 19,887 | 29,779 | +50% | 1 | 1 | 0% | 3,573 | 5,460 | +53% | 0 | 0 | — |
case-13 | fail→fail | 25,084 | 22,986 | -8% | 1 | 1 | 0% | 3,871 | 4,347 | +12% | 0 | 0 | — |
case-14 | fail→pass | 24,834 | 26,728 | +8% | 1 | 1 | 0% | 3,935 | 5,717 | +45% | 0 | 0 | — |
case-15 | fail→fail | 22,200 | 22,836 | +3% | 1 | 1 | 0% | 4,029 | 4,493 | +12% | 0 | 0 | — |
case-16 | fail→pass | 20,964 | 25,478 | +22% | 1 | 1 | 0% | 3,235 | 4,408 | +36% | 0 | 0 | — |
case-17 | fail→pass | 25,607 | 16,031 | -37% | 1 | 1 | 0% | 3,847 | 3,353 | -13% | 0 | 0 | — |
case-18 | fail→fail | 16,939 | 22,748 | +34% | 1 | 1 | 0% | 3,195 | 4,738 | +48% | 0 | 0 | — |
case-19 | fail→pass | 21,751 | 43,004 | +98% | 1 | 1 | 0% | 3,758 | 5,577 | +48% | 0 | 0 | — |
case-20 | fail→fail | 23,976 | 22,758 | -5% | 1 | 1 | 0% | 3,556 | 3,966 | +12% | 0 | 0 | — |
case-21 | fail→pass | 23,304 | 22,565 | -3% | 1 | 1 | 0% | 4,078 | 3,588 | -12% | 0 | 0 | — |
case-22 | fail→fail | 20,147 | 20,461 | +2% | 1 | 1 | 0% | 3,240 | 3,947 | +22% | 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 +50 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.