How does this skill achieve 90%+ token savings?

Instead of sending the full content of every file to the LLM context, the skill executes logic locally and only sends back metadata or a summary of the results.

When should I use the Code Execution skill?

It is best used for operations involving 10 or more files, complex multi-step workflows, or any task where efficiency and performance are primary concerns.

What kind of code analysis can it perform?

It can identify functions, classes, and dependencies locally, returning only the structural metadata so Claude can make decisions without reading every line of code.

Can I use it to automate Git tasks?

Yes, the skill includes a dedicated Git API to handle status checks, adding files, committing changes, and pushing code as part of an automated workflow.

Local Code Execution & Optimization

Name: Local Code Execution & Optimization
Author: pur3v4d3r

bypur3v4d3r

•

Productivity & Workflow

Executes Python scripts locally to perform bulk file operations and complex refactors with up to 99% token savings.

The Code Execution skill empowers Claude to act as a powerful local runner, bypassing the token limits and costs associated with processing large codebases directly in the model context. By providing a suite of specialized APIs for filesystem manipulation, metadata-driven code analysis, and Git integration, it allows users to perform massive refactors, audits, and migrations across hundreds of files while returning only concise summaries to the LLM. This makes it an essential tool for developers managing enterprise-scale repositories who need performance and cost-efficiency.

Key Features

01Integrated Git workflow automation for commits and pushes

02Metadata-only analysis to minimize LLM context usage

03Significant token cost reduction (90-99%) for large-scale operations

041 GitHub stars

05Local Python runtime with specialized filesystem and code analysis APIs

06Automated bulk transformations and identifier renaming across directories

Use Cases

01Auditing codebase complexity and dependency maps without uploading full source code

02Refactoring or renaming variables across repositories with 50+ files

03Automating multi-step migration scripts that require iterative local logic

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add pur3v4d3r/pur3-pkb-codebase code-execution

For use in Claude.ai and ChatGPT

Download Skill