skills.rest
chenhunghan
chenhunghanCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

code-mode

Execute sandboxed scripts against large API responses to reduce context size.

21|5|Updated Mar 9, 2026
Search Tags:#tooling#mcp#sandbox#code-mode#script-execution#context-reduction
One-click install
npx skills add https://github.com/chenhunghan/code-mode-skill --skill code-mode
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: code-mode
Source: https://github.com/chenhunghan/code-mode-skill/tree/main
Command: npx skills add https://github.com/chenhunghan/code-mode-skill --skill code-mode

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reduces context size by enabling a code-mode workflow where LLMs generate small processing scripts that run against large API responses in a sandboxed runtime, ensuring only the script's output enters the context.

Core Features & Use Cases

  • Interactive planning to understand the MCP server and data shapes.
  • Sandbox-selected code execution tool to process raw API responses.
  • Benchmarking to verify context-size reductions before/after.
  • Language-agnostic: supports TypeScript/JavaScript, Python, Go, Rust and more.

Quick Start

Ask your agent to add code mode to an MCP server to start reducing context by executing sandboxed scripts.

Frequently Asked Questions about code-mode

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

FAQPage Schema
How do I reduce context size when processing large API responses in an MCP server?▼

Sandboxed script execution reduces context size by running code-mode scripts against large API responses and returning only the processed output to the MCP server context window.

What is the code-mode workflow for LLMs and when should I use it?▼

Code-mode is a workflow where LLMs generate processing scripts that run against large API responses in a sandboxed runtime. Use it when raw API payloads threaten to exhaust your context window.

Can I execute sandboxed scripts in languages other than Python or JavaScript?▼

Yes, the sandboxed code execution supports TypeScript, JavaScript, Python, Go, and Rust, allowing language-agnostic script execution across various MCP server implementations.

How do I add code mode to an MCP server?▼

Ask your agent to add code mode to an MCP server. Setup requires a SKILL.md frontmatter with name and description, a sandboxed executor to run scripts, and optional references or assets.

Is there a way to benchmark context-size reductions before and after script execution?▼

Yes, benchmarking verifies context-size reductions before and after executing sandboxed scripts, ensuring the code-mode workflow effectively minimizes the payload entering the MCP server context.

Do I need specific dependencies to run sandboxed scripts against API responses?▼

No specific dependencies are required to run sandboxed scripts. You need a SKILL.md frontmatter with name and description, a sandboxed executor, and optional references to guide usage.

Related Skills

Explore More Software Engineering Skills→
glassBead-tcglassBead-tc
5

mcp-code-mode

Consolidate large API surfaces into two-tool MCP patterns for code orchestration.

Community
Advanced
script-reposcript-repo

V3 Performance Optimization

Baseline claude-flow v3 performance and identify bottlenecks across Flash Attention, HNSW indexing, and memory usage.

Community
Advanced
githubgithub
37.6k

copilot-sdk

Embed Copilot agentic workflows into applications using Python, TypeScript, Go, and .NET SDKs.

Official
Advanced
udecodeudecode
434

dev-browser

Automate browser tasks with persistent state and sandboxed JavaScript scripts.

Official
Intermediate