skills.rest
mcfearsome
mcfearsomeCOMMUNITY
Explore all skills by this author →
Website𝕏@mcfearsomeGitHub

code-executor

Execute multi-step MCP tool workflows via disposable subagents running TypeScript or Python.

1|Updated Nov 11, 2025
Search Tags:#mcp#retry logic#code execution#subagent#token optimization#tool orchestration#parallelism
One-click install
npx skills add https://github.com/mcfearsome/cc-mcp-executor-skill --skill code-executor-mcfearsome
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-executor
Source: https://github.com/mcfearsome/cc-mcp-executor-skill/tree/main/code-executor
Command: npx skills add https://github.com/mcfearsome/cc-mcp-executor-skill --skill code-executor-mcfearsome

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Loading many MCP tool schemas into the main Claude Code context causes major token overhead and slows multi-step tool workflows.

Core Features & Use Cases

  • Subagent code execution for tool composition: Offloads multi-tool sequencing to a disposable subagent that writes and runs TypeScript or Python to call MCP tools dynamically.
  • Token-saving workflow orchestration: Keeps the main context free of MCP server/tool schema bloat while still enabling end-to-end multi-tool operations.
  • Built-in patterns for real tasks: Supports common orchestration needs like file batch processing, parallel requests, retry/fallback logic, conditional tool routing, and multi-source aggregation.
  • Use Case: “Read all JSON files in /tmp/data, aggregate the contents, and store results in a database” while keeping the main agent lightweight and focused on orchestration.

Quick Start

Ask Claude Code to “Read all JSON files in /tmp/data, aggregate the data, and store it in database table aggregated_data.”

Frequently Asked Questions about code-executor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I reduce token bloat when orchestrating multi-step MCP tool workflows?▼

Token bloat from MCP tool schemas is reduced by offloading multi-tool sequencing to a disposable subagent that dynamically writes and runs TypeScript or Python to call MCP tools and return summarized results.

Can I run parallel MCP calls and retry fallback logic without loading all tool schemas into the main context?▼

Parallel MCP calls, retry fallback strategies, and conditional tool routing are executed outside the main context by delegating the workflow to a subagent running TypeScript or Python code.

How do I batch process multiple files and aggregate data using MCP tools?▼

Batch file processing and multi-source aggregation are handled by instructing the subagent to read files like JSON datasets, aggregate contents, and store results in a database while keeping the main agent lightweight.

Do I need Deno or Python installed to execute MCP tool orchestration workflows?▼

Yes, the subagent executes workflows by running TypeScript via Deno or Python code locally, utilizing an MCP_CONFIG_PATH to call tools through a local MCP client.

What is the best way to coordinate three or more MCP tool calls together?▼

Coordinating three or more tool calls is best handled by composing a multi-step workflow definition that dispatches execution to a subagent, preventing main context slowdown from schema overhead.

Why does loading multiple MCP servers slow down complex tool workflows?▼

Loading many MCP server schemas causes major token overhead in the main context, which slows multi-step tool workflows and degrades overall orchestration performance.

Related Skills

Explore More Software Engineering Skills→
mcfearsomemcfearsome
1

mcp-executor

Execute multi-tool MCP workflows with retry logic and progress updates.

Community
Advanced
levnikolaevichlevnikolaevich
534

n/a

Coordinate project documentation creation across a multi-phase pipeline.

Community
Advanced
AradotsoAradotso
7

mcp-code-execution-mode

Proxy MCP tool calls through a sandboxed run_python bridge.

Official
Advanced
laitim2001laitim2001

claude-agent-sdk

Orchestrate autonomous Claude agents with MCP tools and safety hooks.

Community
Advanced