Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Add a new tool to the backend OpenAI function calling system. Use when user mentions "new tool", "add tool", "backend function", "agent capability", or wants to extend what the AI agent can do.
.claude/skills/aiskillstore-add-backend-tool/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -49% | 0% |
| case-13 | ✓→✗ | ▼ Worse | -47% | 0% |
backend/main.py to understand existing tool patterns:tools list with function definitionspython def new_tool_name(param1: str, param2: int = 10) -> str: """Docstring explaining the tool.""" try: # Implementation return result except Exception as e: return f"Error: {str(e)}"
tools list:python { "type": "function", "function": { "name": "new_tool_name", "description": "What this tool does and when to use it", "parameters": { "type": "object", "properties": { "param1": {"type": "string", "description": "..."}, "param2": {"type": "integer", "description": "..."} }, "required": ["param1"] } } }
python elif func_name == "new_tool_name": result = new_tool_name(**args)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 12,886 | 18,700 | +45% | 1 | 1 | 0% | 2,677 | 3,795 | +42% | 0 | 0 | — |
case-01 | fail→fail | 26,946 | 10,781 | -60% | 1 | 1 | 0% | 4,083 | 829 | -80% | 0 | 0 | — |
case-02 | fail→fail | 19,447 | 10,565 | -46% | 1 | 1 | 0% | 3,054 | 602 | -80% | 0 | 0 | — |
case-03 | fail→fail | 21,275 | 12,714 | -40% | 1 | 1 | 0% | 3,586 | 830 | -77% | 0 | 0 | — |
case-04 | pass→pass | 18,031 | 32,577 | +81% | 1 | 1 | 0% | 2,556 | 4,342 | +70% | 0 | 0 | — |
case-05 | pass→pass | 22,159 | 20,509 | -7% | 1 | 1 | 0% | 4,140 | 4,343 | +5% | 0 | 0 | — |
case-07 | pass→pass | 21,051 | 18,712 | -11% | 1 | 1 | 0% | 2,750 | 2,868 | +4% | 0 | 0 | — |
case-08 | fail→pass | 18,583 | 12,971 | -30% | 1 | 1 | 0% | 3,025 | 1,845 | -39% | 0 | 0 | — |
case-09 | fail→fail | 27,964 | 8,852 | -68% | 1 | 1 | 0% | 4,275 | 721 | -83% | 0 | 0 | — |
case-10 | pass→pass | 22,557 | 8,423 | -63% | 1 | 1 | 0% | 2,927 | 1,920 | -34% | 0 | 0 | — |
case-11 | pass→pass | 12,345 | 5,748 | -53% | 1 | 1 | 0% | 2,170 | 1,418 | -35% | 0 | 0 | — |
case-12 | pass→fail | 14,960 | 11,444 | -24% | 1 | 1 | 0% | 1,632 | 831 | -49% | 0 | 0 | — |
case-13 | pass→fail | 15,092 | 15,707 | +4% | 1 | 1 | 0% | 1,643 | 878 | -47% | 0 | 0 | — |
case-14 | pass→pass | 5,697 | 11,739 | +106% | 1 | 1 | 0% | 1,011 | 1,216 | +20% | 0 | 0 | — |
case-15 | fail→pass | 18,562 | 12,810 | -31% | 1 | 1 | 0% | 2,224 | 1,743 | -22% | 0 | 0 | — |
case-16 | pass→pass | 17,950 | 9,047 | -50% | 1 | 1 | 0% | 2,468 | 1,247 | -49% | 0 | 0 | — |
case-17 | fail→fail | 15,020 | 8,860 | -41% | 1 | 1 | 0% | 2,581 | 602 | -77% | 0 | 0 | — |
case-18 | pass→pass | 18,741 | 5,559 | -70% | 1 | 1 | 0% | 2,530 | 1,505 | -41% | 0 | 0 | — |
case-19 | fail→pass | 16,863 | 5,631 | -67% | 1 | 1 | 0% | 1,895 | 1,401 | -26% | 0 | 0 | — |
case-20 | pass→pass | 14,568 | 9,692 | -33% | 1 | 1 | 0% | 1,584 | 1,222 | -23% | 0 | 0 | — |
case-21 | pass→pass | 11,860 | 7,461 | -37% | 1 | 1 | 0% | 1,749 | 861 | -51% | 0 | 0 | — |
case-22 | pass→pass | 10,122 | 11,753 | +16% | 1 | 1 | 0% | 1,677 | 1,567 | -7% | 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, and 15 counted toward the lift figure. The other 7 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 +5 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 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.