Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Design tools, scripts, and CLIs that an agent will call, so the interface teaches its own use instead of a wall of prose and examples. Use when building an MCP server or tool definition, writing an agent-facing script, or when an agent keeps misusing a tool it already has.
.claude/skills/neeeophytee-agent-interface-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 22% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 40% | 0% |
Examples teach one path and quietly fence off the others: shown three ways to call a tool, a model tends to produce those three. A well-designed interface teaches the whole space at once. The parameters say what is possible, the description says what is expected, and there is very little left to write.
pending | in_progress | completed teaches the whole state machine without a sentence of prose.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 12,679 | 12,185 | -4% | 1 | 1 | 0% | 1,965 | 2,388 | +22% | 0 | 0 | — |
case-21 | fail→pass | 9,430 | 10,343 | +10% | 1 | 1 | 0% | 1,447 | 2,006 | +39% | 0 | 0 | — |
case-01 | fail→pass | 14,427 | 24,089 | +67% | 1 | 1 | 0% | 2,463 | 3,029 | +23% | 0 | 0 | — |
case-02 | pass→pass | 10,596 | 9,446 | -11% | 1 | 1 | 0% | 1,633 | 1,795 | +10% | 0 | 0 | — |
case-03 | pass→pass | 11,512 | 7,560 | -34% | 1 | 1 | 0% | 1,747 | 1,561 | -11% | 0 | 0 | — |
case-04 | pass→pass | 13,511 | 10,556 | -22% | 1 | 1 | 0% | 2,020 | 2,298 | +14% | 0 | 0 | — |
case-05 | pass→pass | 14,370 | 10,713 | -25% | 1 | 1 | 0% | 2,330 | 2,253 | -3% | 0 | 0 | — |
case-06 | pass→pass | 13,097 | 9,643 | -26% | 1 | 1 | 0% | 2,267 | 2,116 | -7% | 0 | 0 | — |
case-07 | pass→pass | 10,655 | 8,031 | -25% | 1 | 1 | 0% | 1,723 | 1,715 | -0% | 0 | 0 | — |
case-08 | fail→fail | 14,513 | 16,814 | +16% | 1 | 1 | 0% | 2,405 | 3,294 | +37% | 0 | 0 | — |
case-09 | pass→pass | 11,504 | 9,124 | -21% | 1 | 1 | 0% | 1,721 | 1,815 | +5% | 0 | 0 | — |
case-10 | pass→pass | 13,447 | 9,928 | -26% | 1 | 1 | 0% | 2,096 | 1,974 | -6% | 0 | 0 | — |
case-11 | pass→pass | 10,037 | 10,636 | +6% | 1 | 1 | 0% | 1,533 | 2,003 | +31% | 0 | 0 | — |
case-13 | fail→pass | 7,639 | 2,730 | -64% | 1 | 1 | 0% | 1,172 | 942 | -20% | 0 | 0 | — |
case-14 | pass→pass | 11,039 | 10,670 | -3% | 1 | 1 | 0% | 1,668 | 2,146 | +29% | 0 | 0 | — |
case-15 | fail→pass | 9,590 | 10,375 | +8% | 1 | 1 | 0% | 1,462 | 2,040 | +40% | 0 | 0 | — |
case-16 | fail→pass | 12,647 | 9,451 | -25% | 1 | 1 | 0% | 2,335 | 1,977 | -15% | 0 | 0 | — |
case-17 | pass→pass | 15,003 | 9,431 | -37% | 1 | 1 | 0% | 2,577 | 2,119 | -18% | 0 | 0 | — |
case-18 | fail→fail | 19,061 | 6,074 | -68% | 1 | 1 | 0% | 1,544 | 1,453 | -6% | 0 | 0 | — |
case-19 | pass→pass | 15,379 | 10,993 | -29% | 1 | 1 | 0% | 2,406 | 2,322 | -3% | 0 | 0 | — |
case-20 | pass→pass | 8,863 | 7,655 | -14% | 1 | 1 | 0% | 1,565 | 1,758 | +12% | 0 | 0 | — |
case-22 | pass→pass | 11,995 | 6,867 | -43% | 1 | 1 | 0% | 1,843 | 1,550 | -16% | 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 +27 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.