Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →CLI-first tool selection policy for Claude Code. Use when choosing between CLI tools and MCP servers, designing new agents, or reviewing agent tool configurations. Do NOT use for executing commands or running tools -- this is for tool selection decisions during agent/config design only.
.claude/skills/majiayu000-tool-selection/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✓→✗ | ▼ Worse | -87% | 0% |
| case-18 | ✓→✗ | ▼ Worse | -79% | 0% |
| case-09 | ✓→✓ | ＝ Same ✓ | 7% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 22% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | -30% | 0% |
Select the optimal MCP tool by evaluating task complexity, accuracy needs, and performance trade-offs.
Avoid when:
| Task | Load reference | | --- | --- | | Tool selection | skills/tool-selection/references/select.md |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 21,198 | 27,182 | +28% | 1 | 1 | 0% | 2,276 | 2,434 | +7% | 0 | 0 | — |
case-02 | pass→pass | 13,068 | 27,592 | +111% | 1 | 1 | 0% | 1,206 | 1,467 | +22% | 0 | 0 | — |
case-01 | pass→pass | 18,833 | 16,368 | -13% | 1 | 1 | 0% | 2,663 | 1,860 | -30% | 0 | 0 | — |
case-03 | pass→pass | 21,080 | 10,059 | -52% | 1 | 1 | 0% | 3,388 | 1,740 | -49% | 0 | 0 | — |
case-04 | pass→pass | 14,706 | 15,065 | +2% | 1 | 1 | 0% | 1,570 | 1,674 | +7% | 0 | 0 | — |
case-05 | pass→pass | 25,575 | 31,502 | +23% | 1 | 1 | 0% | 2,988 | 3,361 | +12% | 0 | 0 | — |
case-06 | fail→fail | 6,948 | 14,781 | +113% | 1 | 1 | 0% | 922 | 1,162 | +26% | 0 | 0 | — |
case-07 | pass→pass | 15,375 | 12,586 | -18% | 1 | 1 | 0% | 2,377 | 1,269 | -47% | 0 | 0 | — |
case-08 | pass→fail | 19,389 | 17,018 | -12% | 1 | 1 | 0% | 2,750 | 346 | -87% | 0 | 0 | — |
case-10 | fail→fail | 21,345 | 7,386 | -65% | 1 | 1 | 0% | 1,836 | 1,334 | -27% | 0 | 0 | — |
case-11 | pass→pass | 22,103 | 25,875 | +17% | 1 | 1 | 0% | 2,757 | 2,268 | -18% | 0 | 0 | — |
case-12 | pass→pass | 21,333 | 18,948 | -11% | 1 | 1 | 0% | 1,693 | 2,093 | +24% | 0 | 0 | — |
case-13 | pass→pass | 21,657 | 16,176 | -25% | 1 | 1 | 0% | 2,072 | 1,875 | -10% | 0 | 0 | — |
case-14 | pass→pass | 12,091 | 23,786 | +97% | 1 | 1 | 0% | 1,991 | 1,765 | -11% | 0 | 0 | — |
case-15 | pass→pass | 8,278 | 15,109 | +83% | 1 | 1 | 0% | 1,184 | 1,494 | +26% | 0 | 0 | — |
case-16 | pass→pass | 11,117 | 15,878 | +43% | 1 | 1 | 0% | 1,660 | 1,917 | +15% | 0 | 0 | — |
case-17 | pass→pass | 16,825 | 15,330 | -9% | 1 | 1 | 0% | 1,722 | 1,448 | -16% | 0 | 0 | — |
case-18 | pass→fail | 15,627 | 15,262 | -2% | 1 | 1 | 0% | 2,212 | 459 | -79% | 0 | 0 | — |
case-19 | pass→pass | 17,233 | 16,148 | -6% | 1 | 1 | 0% | 1,800 | 1,775 | -1% | 0 | 0 | — |
case-20 | pass→pass | 22,393 | 20,341 | -9% | 1 | 1 | 0% | 2,434 | 2,537 | +4% | 0 | 0 | — |
case-21 | pass→pass | 14,725 | 17,108 | +16% | 1 | 1 | 0% | 2,281 | 1,972 | -14% | 0 | 0 | — |
case-22 | pass→pass | 14,151 | 21,084 | +49% | 1 | 1 | 0% | 2,355 | 1,614 | -31% | 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 20 counted toward the lift figure. The other 2 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 -9 percentage points is the difference between those two pass rates over the 20 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Other measured skills in the registry, with their headline benchmark lift.