Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Add or update workspace MCP servers using holaOS mcp_registry syntax.
.claude/skills/holaboss-ai-mcp-configurator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 147% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 37% | 0% |
Use this skill when the task is to add, remove, or update MCP access in a holaOS workspace.
workspace.yaml at the workspace root. Do not create mcp.json, .cursor/mcp.json, claude_desktop_config.json, or another generic MCP config unless the user explicitly asks for that format.mcp_registry.mcp_registry.allowlist.tool_ids must use strict server.tool syntax.mcp_registry.allowlist.tool_ids or setting it to [] means all discovered tools from enabled configured servers stay available for that run.runtime/harnesses/src/embedded-skills/.Use this shape for remote HTTP MCP servers:
yamlmcp_registry: servers: context7: type: remote url: https://mcp.context7.com/mcp enabled: true timeout_ms: 30000 headers: CONTEXT7_API_KEY: "{env:CONTEXT7_API_KEY}" allowlist: tool_ids: - context7.lookup - context7.search
Header values must be either:
{env:ENV_VAR_NAME}This is invalid:
yamlheaders: CONTEXT7_API_KEY: "{env:ctx7sk-live-abc123}"
That is a placeholder-shaped literal secret, not an environment-variable reference.
Use this shape for local stdio MCP servers:
yamlmcp_registry: servers: my_server: type: local command: - npx - -y - "@acme/my-mcp" enabled: true timeout_ms: 30000 environment: MY_SERVER_API_KEY: "{env:MY_SERVER_API_KEY}"
Local MCP commands must be a non-empty list of command tokens.
Use mcp_registry.catalog for workspace-local tools implemented inside the repo:
yamlmcp_registry: servers: workspace: type: local catalog: workspace.echo: module_path: tools.echo symbol: echo_tool allowlist: tool_ids: - workspace.echo
Notes:
mcp_registry.servers.workspace must stay local.command for workspace.workspace.<tool> allowlist entry must have a matching mcp_registry.catalog entry.If the MCP tools come from a workspace app's app.runtime.yaml mcp.tools, prefer the app-managed flow:
mcp_registry.servers.<app_id> and mcp_registry.allowlist.tool_idsworkspace.yaml and confirm the final shape is mcp_registry, not a generic MCP config format.server.tool.mcp_registry.servers.{env:ENV_VAR_NAME} when environment indirection is intended.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,587 | 7,093 | +8% | 1 | 1 | 0% | 1,158 | 1,219 | +5% | 0 | 0 | — |
case-02 | fail→fail | 10,231 | 5,945 | -42% | 1 | 1 | 0% | 1,620 | 1,112 | -31% | 0 | 0 | — |
case-03 | fail→fail | 11,142 | 5,088 | -54% | 1 | 1 | 0% | 1,627 | 1,178 | -28% | 0 | 0 | — |
case-04 | fail→pass | 9,141 | 4,344 | -52% | 1 | 1 | 0% | 1,297 | 1,861 | +43% | 0 | 0 | — |
case-05 | fail→fail | 5,012 | 4,964 | -1% | 1 | 1 | 0% | 169 | 1,376 | +714% | 0 | 0 | — |
case-06 | fail→pass | 10,009 | 8,538 | -15% | 1 | 1 | 0% | 1,761 | 2,296 | +30% | 0 | 0 | — |
case-07 | fail→pass | 3,139 | 3,428 | +9% | 1 | 1 | 0% | 612 | 1,512 | +147% | 0 | 0 | — |
case-08 | pass→pass | 9,660 | 4,554 | -53% | 1 | 1 | 0% | 1,478 | 1,853 | +25% | 0 | 0 | — |
case-09 | fail→pass | 7,313 | 4,203 | -43% | 1 | 1 | 0% | 1,270 | 1,748 | +38% | 0 | 0 | — |
case-10 | pass→pass | 7,303 | 4,226 | -42% | 1 | 1 | 0% | 1,181 | 1,563 | +32% | 0 | 0 | — |
case-11 | pass→pass | 3,688 | 4,271 | +16% | 1 | 1 | 0% | 730 | 1,588 | +118% | 0 | 0 | — |
case-12 | pass→pass | 6,190 | 4,722 | -24% | 1 | 1 | 0% | 1,001 | 1,920 | +92% | 0 | 0 | — |
case-13 | fail→pass | 7,550 | 5,341 | -29% | 1 | 1 | 0% | 1,251 | 1,712 | +37% | 0 | 0 | — |
case-14 | fail→fail | 17,443 | 5,710 | -67% | 1 | 1 | 0% | 929 | 1,310 | +41% | 0 | 0 | — |
case-15 | fail→pass | 11,846 | 3,699 | -69% | 1 | 1 | 0% | 1,744 | 1,652 | -5% | 0 | 0 | — |
case-16 | fail→pass | 8,415 | 2,801 | -67% | 1 | 1 | 0% | 1,150 | 1,400 | +22% | 0 | 0 | — |
case-17 | fail→pass | 4,542 | 3,802 | -16% | 1 | 1 | 0% | 808 | 1,712 | +112% | 0 | 0 | — |
case-18 | pass→pass | 9,137 | 2,548 | -72% | 1 | 1 | 0% | 1,223 | 1,266 | +4% | 0 | 0 | — |
case-19 | fail→pass | 13,231 | 10,843 | -18% | 1 | 1 | 0% | 2,179 | 2,573 | +18% | 0 | 0 | — |
case-20 | fail→pass | 6,801 | 2,853 | -58% | 1 | 1 | 0% | 1,017 | 1,470 | +45% | 0 | 0 | — |
case-21 | pass→pass | 7,689 | 5,275 | -31% | 1 | 1 | 0% | 1,366 | 1,857 | +36% | 0 | 0 | — |
case-22 | fail→pass | 4,655 | 2,974 | -36% | 1 | 1 | 0% | 874 | 1,514 | +73% | 0 | 0 | — |
case-23 | pass→pass | 6,783 | 4,209 | -38% | 1 | 1 | 0% | 989 | 1,788 | +81% | 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. 23 cases were attempted, and 19 counted toward the lift figure. The other 4 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 +48 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.