$ show fumiya-kume/claude-code
This repository provides a collection of Claude Code plugins designed to enhance AI-assisted development workflows. It consists of four specialized utilities: deslop for removing AI-generated code artifacts, dig for conducting deep exploratory interviews to clarify plans, fix-ci for automatically diagnosing and resolving CI failures, and decomposition for breaking down complex tasks into detailed, actionable todos. Each plugin implements sophisticated patterns for improving developer productivity when working with AI assistants. The system follows a modular architecture where each plugin is self-contained with its own command implementation, skill definitions, and metadata. Plugins can be installed via Claude Code's plugin marketplace system and integrate directly into the AI assistant's workflow. The plugins leverage Claude Code's built-in tools (Read, Edit, Write, Bash, etc.) to interact with the development environment, analyze code, and make targeted improvements. Rather than being a monolithic application, it offers composable tools that address specific pain points in AI-assisted development.
The LLM identified 4 files worth highlighting for reuse:
Implements a systematic workflow for identifying and removing AI-generated code slop from git diffs. The component analyzes changes against a base branch, detects patterns indicative of AI-generated code (excessive comments, abnormal defensive checks, inconsistent styles), and removes them while preserving intended behavior. Demonstrates the AI-assisted code review pattern where automated tools help maintain code quality in AI-generated contributions.
Provides a framework for deep exploratory questioning that surfaces hidden assumptions and risks in plans. It implements assumption mapping by risk level, iterative deep questioning with 2-3 focused questions per round, and structured documentation of discoveries. The component demonstrates the investigative interviewing pattern where AI acts as a skilled interviewer to uncover implicit assumptions through carefully structured, progressive questioning.
Delivers an automated CI failure diagnosis and resolution system that uses GitHub CLI to check status, retrieve failure logs, identify root causes, and implement fixes. The component follows a precise workflow: authenticate, check status, inspect logs, confirm root cause, explain and propose fix, implement fix, commit/push, and verify. Demonstrates the automated DevOps pattern where AI agents can systematically troubleshoot and resolve infrastructure issues.
Implements a comprehensive task breakdown system that converts complex goals into specific, achievable todos with rich descriptions (What, Where, How, Why, Verify). The component explores the codebase, identifies major components, optionally interviews for unclear points, creates detailed todos following strict quality criteria, and writes them to the todo system. Demonstrates the hierarchical task decomposition pattern where AI assists in breaking down work into executable units with clear acceptance criteria.