Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Manage the Sandboxed.sh library (skills, agents, commands, tools, rules, MCPs) via Library API tools. Trigger terms: library, skill, agent, command, tool, rule, MCP, save skill, create skill.
.claude/skills/th0rgal-library-management/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 42% | 0% |
The Sandboxed.sh Library is a Git-backed configuration repo that stores reusable skills, agents, commands, tools, rules, MCP servers, and workspace templates. Use the library-* tools to read and update that repo.
Tool names follow the pattern <filename>_<export>.
library-skills.ts)library-skills_list_skillslibrary-skills_get_skilllibrary-skills_save_skilllibrary-skills_delete_skilllibrary-agents.ts)library-agents_list_agentslibrary-agents_get_agentlibrary-agents_save_agentlibrary-agents_delete_agentlibrary-commands.ts)library-commands_list_commands, library-commands_get_command, library-commands_save_command, library-commands_delete_commandlibrary-commands_list_tools, library-commands_get_tool, library-commands_save_tool, library-commands_delete_toollibrary-commands_list_rules, library-commands_get_rule, library-commands_save_rule, library-commands_delete_rulelibrary-git.ts)library-git_get_mcps, library-git_save_mcpslibrary-git_status, library-git_sync, library-git_commit, library-git_pushskill/<name>/SKILL.md)yaml--- name: skill-name description: What this skill does --- Instructions for using this skill...
agent/<name>.md)yaml--- description: Agent description mode: primary | subagent model: provider/model-id hidden: true | false color: "#44BA81" tools: "*": false "read": true "write": true permission: edit: ask | allow | deny bash: "*": ask rules: - rule-name --- Agent system prompt...
command/<name>.md)yaml--- description: Command description model: provider/model-id subtask: true | false agent: agent-name --- Command prompt template. Use $ARGUMENTS for user input.
tool/<name>.ts)typescriptimport { tool } from "@opencode-ai/plugin" export const my_tool = tool({ description: "What it does", args: { param: tool.schema.string().describe("Param description") }, async execute(args) { return "result" }, })
rule/<name>.md)yaml--- description: Rule description --- Rule instructions applied to agents referencing this rule.
mcp/servers.json)json{ "server-name": { "type": "local", "command": ["npx", "package-name"], "env": { "KEY": "value" }, "enabled": true }, "remote-server": { "type": "remote", "url": "https://mcp.example.com", "headers": { "Authorization": "Bearer token" }, "enabled": true } }
library-git_status before pushing| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 1,633 | 3,687 | +126% | 1 | 1 | 0% | 236 | 1,290 | +447% | 0 | 0 | — |
case-01 | fail→fail | 12,050 | 5,295 | -56% | 1 | 1 | 0% | 267 | 1,426 | +434% | 0 | 0 | — |
case-02 | fail→fail | 17,979 | 5,141 | -71% | 1 | 1 | 0% | 2,310 | 1,282 | -45% | 0 | 0 | — |
case-04 | fail→fail | 4,500 | 2,793 | -38% | 1 | 1 | 0% | 663 | 1,481 | +123% | 0 | 0 | — |
case-05 | fail→fail | 9,340 | 2,948 | -68% | 1 | 1 | 0% | 1,278 | 1,474 | +15% | 0 | 0 | — |
case-06 | fail→pass | 11,916 | 2,470 | -79% | 1 | 1 | 0% | 1,924 | 1,409 | -27% | 0 | 0 | — |
case-07 | fail→pass | 7,670 | 2,951 | -62% | 1 | 1 | 0% | 1,036 | 1,544 | +49% | 0 | 0 | — |
case-08 | fail→fail | 3,521 | 1,615 | -54% | 1 | 1 | 0% | 520 | 1,272 | +145% | 0 | 0 | — |
case-09 | pass→pass | 7,362 | 2,883 | -61% | 1 | 1 | 0% | 1,179 | 1,505 | +28% | 0 | 0 | — |
case-10 | pass→pass | 8,648 | 1,830 | -79% | 1 | 1 | 0% | 1,350 | 1,253 | -7% | 0 | 0 | — |
case-11 | fail→pass | 13,826 | 2,342 | -83% | 1 | 1 | 0% | 2,112 | 1,399 | -34% | 0 | 0 | — |
case-12 | fail→pass | 10,749 | 3,763 | -65% | 1 | 1 | 0% | 1,771 | 1,595 | -10% | 0 | 0 | — |
case-13 | pass→pass | 5,214 | 2,717 | -48% | 1 | 1 | 0% | 878 | 1,411 | +61% | 0 | 0 | — |
case-14 | pass→pass | 7,363 | 1,760 | -76% | 1 | 1 | 0% | 1,166 | 1,370 | +17% | 0 | 0 | — |
case-15 | pass→pass | 12,094 | 3,765 | -69% | 1 | 1 | 0% | 2,099 | 1,622 | -23% | 0 | 0 | — |
case-16 | pass→pass | 8,245 | 5,089 | -38% | 1 | 1 | 0% | 1,284 | 1,839 | +43% | 0 | 0 | — |
case-17 | pass→pass | 9,553 | 3,263 | -66% | 1 | 1 | 0% | 1,565 | 1,552 | -1% | 0 | 0 | — |
case-18 | fail→pass | 6,247 | 1,948 | -69% | 1 | 1 | 0% | 924 | 1,311 | +42% | 0 | 0 | — |
case-19 | pass→pass | 4,121 | 2,868 | -30% | 1 | 1 | 0% | 607 | 1,419 | +134% | 0 | 0 | — |
case-20 | pass→pass | 5,667 | 4,167 | -26% | 1 | 1 | 0% | 917 | 1,674 | +83% | 0 | 0 | — |
case-21 | pass→pass | 3,641 | 4,504 | +24% | 1 | 1 | 0% | 418 | 1,734 | +315% | 0 | 0 | — |
case-22 | pass→pass | 5,135 | 2,633 | -49% | 1 | 1 | 0% | 844 | 1,427 | +69% | 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 19 counted toward the lift figure. The other 3 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 +23 percentage points is the difference between those two pass rates over the 19 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.