Does it integrate with other Claude Code commands?

Yes, after a research session is complete, it suggests logical next steps such as using /think for architectural design or /fix for implementation.

What tools does it use for investigation?

It leverages specialized sub-agents for architecture and data flow, semantic search via yomu, and standard terminal utilities like grep, git logs, and tree.

What is the main purpose of the Research skill?

The Research skill is designed to perform technical deep-dives into a codebase to answer questions, investigate bugs, or plan features without actually modifying any files.

How does it verify its findings?

It uses a 'Strong Inference' protocol for bugs and applies verifiability markers to every finding to distinguish between facts, probable hypotheses, and unknown areas.

Where are the research results stored?

All findings are automatically synthesized into a timestamped Markdown file and saved in your project's 'workspace/research/' directory for future reference.

Codebase Research & Investigation

Name: Codebase Research & Investigation
Author: thkt

bythkt

•

开发者工具

Conducts deep technical analysis and codebase investigations to provide evidence-based insights without modifying code.

The Research skill is a specialized diagnostic agent for Claude Code designed to explore and document complex codebases. It automates the discovery process by deploying intent-aware analyzers for architecture, data models, and API structures while performing semantic searches to find conceptually related code. By following a structured investigation protocol—including hypothesis testing for bug analysis—it synthesizes findings into verifiable reports, allowing developers to plan features or debug issues with full context before writing a single line of code.

主要功能

01Verifiable finding categorization using confidence-based markers ([✓]/[→]/[?])

02Automated bug investigation protocol with hypothesis testing and evidence markers

03Semantic code search using yomu to find conceptually related implementation patterns

04Structured Markdown output saved automatically to the workspace research directory

058 GitHub stars

06Intent-aware parallel analysis of architecture, data flow, and domain logic

使用场景

01Investigating the root cause of complex bugs before attempting a code fix

02Mapping out unfamiliar codebases to understand execution paths and data models

03Conducting technical feasibility studies and domain analysis for new feature planning

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add thkt/dotclaude research

For use in Claude.ai and ChatGPT