Claude-skill-registry-data mcp-usage-orchestrator

Enforce calling mcp__sequential-thinking__sequentialthinking for every user request and verify configured MCP servers (via .mcp.json or MCP resource listing) to route tasks to the right MCP tools. Use for all tasks in this project, especially when selecting MCP tools, confirming installed MCP servers, orchestrating parallel MCP calls, or choosing between playwright-test and github-fetcher.

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry-data
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry-data "$T" && mkdir -p ~/.claude/skills && cp -r "$T/data/mcp-usage-orchestrator" ~/.claude/skills/majiayu000-claude-skill-registry-data-mcp-usage-orchestrator && rm -rf "$T"
manifest: data/mcp-usage-orchestrator/SKILL.md
source content

MCP Usage Orchestrator

Goal

Use sequential-thinking on every request and keep MCP tool usage aligned with the MCP servers configured in this repo.

Mandatory Per-Task Steps

  1. Call
    mcp__sequential-thinking__sequentialthinking
    at the start of every user request. Keep the thoughts concise and action-oriented, even for simple tasks.
  2. If this conversation has not yet verified MCP configuration, read
    .mcp.json
    at the repo root and list the configured
    mcpServers
    keys. Re-check when the repo changes or the user mentions MCP setup.
    • If
      .mcp.json
      is missing or unclear, call
      list_mcp_resources
      to confirm reachable servers and note any gaps.
  3. Use MCP tools that match the configured servers and the task intent. Prefer the most direct MCP tool for the job and avoid unnecessary calls.

Tool Routing Heuristics

  • Browser automation, UI flows, or Playwright tests: use
    mcp__playwright-test__*
    .
  • Web page content retrieval: use
    mcp__github-fetcher__*
    .
  • Independent tool calls that can run concurrently: use
    multi_tool_use.parallel
    .
  • If a task requires an MCP server not in
    .mcp.json
    , say so and ask the user whether to install/configure it.

Notes

  • Keep the MCP inventory in working memory once checked; do not repeatedly read
    .mcp.json
    unless needed.
  • If MCP configuration is missing or invalid, state the problem and proceed with non-MCP tools as appropriate.