Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Discovers and executes MCP tools on-demand via agentsbox_* tools. Use when you need to find the right tool, when a user asks to use a tool you don’t have in context, or when troubleshooting MCP/tool availability.
.claude/skills/dicklesworthstone-agentsbox/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -24% | 0% |
Minimize context bloat by using a small, stable tool surface:
agentsbox_search_bm25 (natural language)agentsbox_search_regex (pattern search)agentsbox_execute (run discovered tool)agentsbox_status (health/status)agentsbox_perf (performance)agentsbox_test (tool verification)agentsbox_search_bm25.agentsbox_search_regex.schema and required fields.agentsbox_execute({ toolId, arguments }).toolId format: {serverName}_{toolName}.arguments is a JSON string (or omit / {} when none required).agentsbox_status({}) and include server diagnostics.1) Search:
textagentsbox_search_bm25({ "text": "search the web for information", "limit": 5 })
2) Execute:
textagentsbox_execute({ "toolId": "tavily_tavily_search", "arguments": "{\"query\":\"latest AI news\",\"max_results\":5}" })
agentsbox_status will show zero configured servers."server_.*".| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,227 | 7,825 | +8% | 1 | 1 | 0% | 1,152 | 681 | -41% | 0 | 0 | — |
case-02 | fail→fail | 6,263 | 2,802 | -55% | 1 | 1 | 0% | 553 | 686 | +24% | 0 | 0 | — |
case-03 | fail→fail | 24,646 | 5,813 | -76% | 1 | 1 | 0% | 2,566 | 695 | -73% | 0 | 0 | — |
case-04 | fail→pass | 12,171 | 6,893 | -43% | 1 | 1 | 0% | 1,711 | 904 | -47% | 0 | 0 | — |
case-19 | fail→fail | 13,330 | 2,888 | -78% | 1 | 1 | 0% | 2,083 | 909 | -56% | 0 | 0 | — |
case-05 | fail→fail | 10,257 | 9,952 | -3% | 1 | 1 | 0% | 1,667 | 756 | -55% | 0 | 0 | — |
case-06 | fail→pass | 7,580 | 3,871 | -49% | 1 | 1 | 0% | 1,360 | 1,076 | -21% | 0 | 0 | — |
case-07 | fail→pass | 3,797 | 2,475 | -35% | 1 | 1 | 0% | 626 | 885 | +41% | 0 | 0 | — |
case-08 | fail→pass | 8,171 | 2,276 | -72% | 1 | 1 | 0% | 1,279 | 822 | -36% | 0 | 0 | — |
case-09 | fail→pass | 7,990 | 3,090 | -61% | 1 | 1 | 0% | 1,251 | 955 | -24% | 0 | 0 | — |
case-10 | fail→pass | 11,385 | 2,042 | -82% | 1 | 1 | 0% | 1,683 | 790 | -53% | 0 | 0 | — |
case-11 | fail→pass | 5,617 | 2,422 | -57% | 1 | 1 | 0% | 767 | 834 | +9% | 0 | 0 | — |
case-12 | pass→pass | 4,647 | 3,900 | -16% | 1 | 1 | 0% | 635 | 1,117 | +76% | 0 | 0 | — |
case-13 | pass→pass | 9,214 | 3,476 | -62% | 1 | 1 | 0% | 1,406 | 1,087 | -23% | 0 | 0 | — |
case-14 | fail→pass | 11,142 | 3,878 | -65% | 1 | 1 | 0% | 1,561 | 914 | -41% | 0 | 0 | — |
case-15 | fail→pass | 7,741 | 2,129 | -72% | 1 | 1 | 0% | 1,325 | 730 | -45% | 0 | 0 | — |
case-16 | pass→pass | 9,471 | 10,118 | +7% | 1 | 1 | 0% | 1,433 | 946 | -34% | 0 | 0 | — |
case-17 | fail→pass | 8,287 | 7,700 | -7% | 1 | 1 | 0% | 1,459 | 971 | -33% | 0 | 0 | — |
case-18 | fail→pass | 8,518 | 4,066 | -52% | 1 | 1 | 0% | 1,469 | 747 | -49% | 0 | 0 | — |
case-20 | pass→fail | 18,235 | 7,637 | -58% | 1 | 1 | 0% | 3,201 | 722 | -77% | 0 | 0 | — |
case-21 | pass→pass | 8,940 | 6,408 | -28% | 1 | 1 | 0% | 1,786 | 1,841 | +3% | 0 | 0 | — |
case-22 | pass→pass | 15,381 | 10,972 | -29% | 1 | 1 | 0% | 2,404 | 2,707 | +13% | 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 17 counted toward the lift figure. The other 5 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 +45 percentage points is the difference between those two pass rates over the 17 comparable cases. 1 case got worse with the skill loaded, and it is 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.