Prescriptive and Contextual Technical Debt Management with LLM and SonarQube

This record contains the full research artefact for the paper 
"Prescriptive and Contextual Technical Debt Management with LLM and SonarQube".

The artefact enables reproduction of the proposed LLM-centric technical debt
management pipeline integrated with GitHub Pull Requests and SonarCloud.

Contents:
- GitHub Actions workflow for PR-level analysis
- SonarCloud issue extraction and prioritization logic
- LLM-based Technical Debt Analyzer agent (Flowise AgentFlow)
- LLM-as-a-Judge evaluation agent following Rule2Text principles
- SQL queries used to extract and filter the Technical Debt Dataset v2.0

The system augments SonarCloud’s diagnostic messages with contextual and
prescriptive explanations generated by Large Language Models (LLMs).
Evaluation is performed using an automated LLM-as-a-Judge framework that
scores contextual relevance and prescriptive depth on a 1–5 Likert scale.

Target projects:
- Apache Commons BCEL
- Apache Commons Collections

The artefact corresponds to Sections III–IV of the paper and supports
reproducibility of all experiments described therein.

Requirements:
- GitHub account
- SonarCloud account
- Flowise (cloud or self-hosted)
- OpenAI-compatible API key (Analyzer agent)
- Gemini or equivalent LLM API key (Judge agent)

Full reproduction instructions are provided in README.md.