mcp-cli_skill

This skill enables on-demand MCP server discovery and calls using the mcp CLI, helping you explore capabilities without permanent integrations.
  • Shell

182

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill obra/superpowers-lab --skill mcp-cli

  • SKILL.md9.0 KB

Overview

This skill provides an on-demand workflow for interacting with MCP servers via the mcp CLI. It lets you discover available tools, resources, and prompts and invoke server capabilities without creating permanent integrations. Use it to test, debug, or make ad-hoc calls while keeping the agent context clean.

How this skill works

The skill runs the mcp CLI to query a target server command or URL and lists available tools, resources, and prompt templates. It supports multiple transport types (stdio, HTTP, SSE) and can show detailed JSON schemas for each tool. Calls are executed with mcp call using JSON parameters and optional output formatting for human or programmatic consumption.

When to use it

  • Explore a server’s capabilities before committing to an integration
  • Make one-off tool calls without polluting agent context
  • Debug or test MCP servers during development
  • Access servers that aren’t pre-configured or hosted remotely
  • Perform quick resource reads or prompt retrievals

Best practices

  • Always run mcp tools first to inspect available tool names and parameter schemas
  • Use -f json when you need machine-readable output for scripting or jq processing
  • Validate parameter types exactly (string, number, boolean, arrays, optional flags) before calling
  • Create temporary aliases for repeated multi-step flows and remove them when done
  • Limit server capabilities with mcp guard to reduce risk when exposing sensitive operations

Example use cases

  • List filesystem tools and read a README via a temporary filesystem server: mcp tools ... then mcp call read_file
  • Query a memory/knowledge-graph server: discover create_entities and search_nodes then call search_nodes with a query
  • Invoke a GitHub MCP server in Docker with a token to list repositories or read PR metadata
  • Retrieve and fill a prompt template from an MCP prompt server, then call a processing tool with the returned prompt
  • Debug startup or parameter issues by requesting pretty JSON schemas and enabling server logs

FAQ

Install mcp at ~/.local/bin/mcp and ensure PATH includes $HOME/.local/bin.

How do I avoid JSON parse errors in shell parameters?

Wrap JSON in single quotes, escape special characters, or supply --params "$(cat params.json)" to read from a file.

Can I restrict what a server can do during a session?

Yes. Use mcp guard with allow/deny patterns to limit tool access, e.g. allow only read_* and list_* operations.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational