convergence-research

Generate a compact, fact-only research document from a codebase.

2|Updated Mar 31, 2026

Search Tags:#automation #documentation #research #codebase #analysis #agents

Ready to run

npx skills add https://github.com/c-sonnier/convergence --skill convergence-research

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: convergence-research
Source: https://github.com/c-sonnier/convergence/tree/main/plugins/convergence/skills/research
Command: npx skills add https://github.com/c-sonnier/convergence --skill convergence-research

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill captures codebase understanding in a compact, fact-only report, enabling fast onboarding and informed decision-making for feature design.

Core Features & Use Cases

Question-driven research: Generates targeted questions before researching and uses ticket-blind exploration.
Parallel agent dispatch: Spawns multiple agents in parallel to cover codebase areas.
Compact, audit-ready output: Produces a concise document (under 500 lines) with file paths, signatures, patterns, and history.
Git history context: Includes recent git history for touched files to preserve context.

Quick Start

Invoke convergence-research against a target repository to generate a compact, fact-only research document.

Frequently Asked Questions about convergence-research

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

FAQPage Schema

How do I generate a concise research document from my codebase?▼

To generate a concise research document from your codebase, invoke the research tool against a target repository. It dispatches parallel agents to explore the codebase and produces a compact, fact-only report under 500 lines for fast onboarding.

What is question-driven codebase analysis and how does it work?▼

Question-driven codebase analysis generates targeted questions before researching a repository. It uses ticket-blind exploration and spawns multiple parallel agents to cover different codebase areas, aggregating findings into a concise, audit-ready document.

Can I use parallel agents to analyze a large repository without hitting output limits?▼

Yes, you can use parallel agents to analyze a large repository while keeping output manageable. The system dispatches multiple agents concurrently to cover different codebase areas and strictly limits the final aggregated research document to under 500 lines.

What is the best way to capture codebase insights for fast onboarding?▼

The best way to capture codebase insights for fast onboarding is to generate a compact, fact-only research document. This approach includes file paths, signatures, patterns, and recent git history for touched files, ensuring informed decision-making for feature design.

Does the codebase research output include git history context for touched files?▼

Yes, the codebase research output includes git history context. It records recent git history for touched files within the compact, audit-ready document to preserve context and provide a deeper understanding of codebase evolution.

Why does the generated research document avoid opinions and subjective analysis?▼

The generated research document avoids opinions to provide a fact-only report suitable for audit and onboarding. It strictly captures file paths, signatures, patterns, and git history, ensuring objective codebase understanding for informed feature design decisions.