Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tools are how AI agents interact with the world. A well-designed tool is the difference between an agent that works and one that hallucinates, fails silently, or costs 10x more tokens than necessary. This skill covers tool design from schema to error handling. JSON Schema best practices, description writing that actually helps the LLM, validation, and the emerging MCP standard that's becoming the lingua franca for AI tools. Key insight: Tool descriptions are more important than tool implementa
.claude/skills/dokhacgiakhoa-agent-tool-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✓→✓ | ＝ Same ✓ | 11% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -10% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 2% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -4% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | 44% | 0% |
You are an expert in the interface between LLMs and the outside world. You've seen tools that work beautifully and tools that cause agents to hallucinate, loop, or fail silently. The difference is almost always in the design, not the implementation.
Your core insight: The LLM never sees your code. It only sees the schema and description. A perfectly implemented tool with a vague description will fail. A simple tool with crystal-clear documentation will succeed.
You push for explicit error hand
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 10,193 | 14,251 | +40% | 1 | 1 | 0% | 1,728 | 1,913 | +11% | 0 | 0 | — |
case-02 | pass→pass | 13,961 | 11,649 | -17% | 1 | 1 | 0% | 2,535 | 2,290 | -10% | 0 | 0 | — |
case-03 | pass→pass | 13,950 | 12,417 | -11% | 1 | 1 | 0% | 2,378 | 2,427 | +2% | 0 | 0 | — |
case-04 | pass→pass | 15,273 | 13,350 | -13% | 1 | 1 | 0% | 3,019 | 2,900 | -4% | 0 | 0 | — |
case-05 | pass→pass | 17,857 | 21,166 | +19% | 1 | 1 | 0% | 3,271 | 4,704 | +44% | 0 | 0 | — |
case-06 | pass→pass | 16,782 | 22,572 | +35% | 1 | 1 | 0% | 3,699 | 4,931 | +33% | 0 | 0 | — |
case-07 | pass→pass | 3,331 | 4,313 | +29% | 1 | 1 | 0% | 679 | 1,070 | +58% | 0 | 0 | — |
case-08 | fail→fail | 7,755 | 7,789 | +0% | 1 | 1 | 0% | 1,482 | 1,882 | +27% | 0 | 0 | — |
case-09 | fail→fail | 7,580 | 7,762 | +2% | 1 | 1 | 0% | 1,504 | 1,894 | +26% | 0 | 0 | — |
case-10 | pass→pass | 16,106 | 17,526 | +9% | 1 | 1 | 0% | 3,074 | 3,544 | +15% | 0 | 0 | — |
case-11 | pass→pass | 8,046 | 10,884 | +35% | 1 | 1 | 0% | 1,497 | 1,718 | +15% | 0 | 0 | — |
case-12 | pass→pass | 21,409 | 27,479 | +28% | 1 | 1 | 0% | 4,528 | 5,882 | +30% | 0 | 0 | — |
case-13 | pass→pass | 15,762 | 10,523 | -33% | 1 | 1 | 0% | 2,384 | 2,310 | -3% | 0 | 0 | — |
case-14 | pass→pass | 15,846 | 15,197 | -4% | 1 | 1 | 0% | 3,145 | 3,476 | +11% | 0 | 0 | — |
case-15 | pass→pass | 9,076 | 10,535 | +16% | 1 | 1 | 0% | 1,611 | 2,170 | +35% | 0 | 0 | — |
case-16 | pass→pass | 12,598 | 15,374 | +22% | 1 | 1 | 0% | 2,279 | 2,608 | +14% | 0 | 0 | — |
case-17 | pass→pass | 6,204 | 7,892 | +27% | 1 | 1 | 0% | 1,151 | 1,789 | +55% | 0 | 0 | — |
case-18 | pass→pass | 12,619 | 11,208 | -11% | 1 | 1 | 0% | 2,304 | 2,249 | -2% | 0 | 0 | — |
case-19 | pass→pass | 10,348 | 11,747 | +14% | 1 | 1 | 0% | 2,159 | 2,668 | +24% | 0 | 0 | — |
case-20 | pass→pass | 12,104 | 9,343 | -23% | 1 | 1 | 0% | 2,296 | 2,140 | -7% | 0 | 0 | — |
case-21 | pass→pass | 10,936 | 14,501 | +33% | 1 | 1 | 0% | 1,960 | 2,777 | +42% | 0 | 0 | — |
case-22 | pass→pass | 13,024 | 9,739 | -25% | 1 | 1 | 0% | 2,439 | 2,277 | -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. The headline lift of 0 percentage points is the difference between those two pass rates over the 22 comparable cases.
Other measured skills in the registry, with their headline benchmark lift.