mcp-servers_skill

Overview

This skill pilots the MCP server reference and provides concise API details, parameters, and usage examples for each Pilot MCP tool. It centralizes workflows for persistent memory, library docs, web search/fetch, GitHub code search, and structural code analysis. Use it to discover tools with ToolSearch and call them directly with the exact parameter patterns shown. The skill focuses on practical examples and when-to-use guidance for faster, accurate tool selection.

How this skill works

The skill documents how to discover MCP tools via ToolSearch and invoke them by their mcp__plugin_pilot_... prefix. For each server it lists the core tools, required parameters, recommended multi-step workflows, and short code snippets that demonstrate typical calls. It also highlights rate limits and call-budget guidance (for example, mem-search’s 3-step token-efficient workflow and context7’s 2-step pattern).

When to use it

• You need persistent project memory or to search past decisions (mem-search).
• You want up-to-date library or framework docs and examples (context7).
• You need web results, README extraction, or full rendered pages (web-search / web-fetch).
• You want real-world production code examples from public repos (grep-mcp).
• You need structural code analysis, call tracing, or blast-radius impact (codebase-memory-mcp).
• You want quick ToolSearch query examples and exact parameter references for MCP tools.

Best practices

• Always run ToolSearch with an appropriate prefix (e.g., +mem-search) before calling a tool to load it lazily.
• For mem-search follow the 3-step workflow: search → timeline → get_observations, fetching full details only for selected IDs.
• Limit calls per tool: e.g., context7 allows up to 3 calls per question; use descriptive queries to reduce back-and-forth.
• Use grep-mcp’s regex mode with (?s) prefix for multiline matching and set matchCase when needed.
• Index a repository first in codebase-memory-mcp and check index_status before running structural queries or trace_call_path.

Example use cases

• Recover design decisions: mem-search search for a feature, timeline for context, then get_observations for full notes.
• Find how a third-party library implements fixtures: context7 resolve-library-id then query-docs with focused queries.
• Locate a fragile code pattern in the wild: grep-mcp searchGitHub with a literal or regex query and language filters.
• Fetch a JavaScript-heavy docs page: web-fetch fetch_url with waitUntil=networkidle and returnHtml to get rendered content.
• Assess change impact: codebase-memory-mcp detect_changes on a git diff, then trace_call_path to enumerate affected callers and callees.