Lore: Repurposing Git Commit Messages as a Structured Knowledge Protocol for AI Coding Agents

TL;DR

Lore protocol repurposes git commit messages into structured knowledge using git trailers, enhancing decision records for AI coding agents.

cs.SE 🔴 Advanced 2026-03-17 54 views
Ivan Stetsenko
AI coding git knowledge management protocol software engineering

Key Findings

Methodology

The Lore protocol utilizes git trailers to transform commit messages into self-contained decision records, including constraints, rejected alternatives, agent directives, and verification metadata. This protocol requires no additional infrastructure beyond git and is queryable via a CLI tool. Lore addresses the gap in capturing implementation-level decisions that traditional Architecture Decision Records (ADRs) miss by embedding knowledge directly into commits.

Key Results

  • Result 1: Lore demonstrated effectiveness in experiments by significantly improving decision efficiency and accuracy for AI coding agents through structured knowledge queries in commit messages.
  • Result 2: Compared to five competing methods, Lore offers a lighter-weight knowledge management solution without requiring additional infrastructure.
  • Result 3: In comparative experiments, Lore showed higher synchronicity and practicality in handling implementation-level decisions compared to traditional ADRs.

Significance

The introduction of the Lore protocol represents a significant shift in software engineering, particularly as AI coding agents become primary producers and consumers of code. By transforming commit messages into structured knowledge records, Lore not only addresses the issue of knowledge loss in traditional methods but also offers new possibilities for future code maintenance and inter-agent knowledge transfer. Its lightweight nature makes it easy to implement across various projects, especially those with limited resources.

Technical Contribution

Lore's technical contributions lie in its innovative use of git trailers for knowledge recording, avoiding the synchronization issues of traditional ADRs and providing more granular decision records at the implementation level. Compared to existing knowledge graphs and version-controlled agent memory systems, Lore requires no complex infrastructure, reducing implementation costs. Additionally, the CLI tool simplifies knowledge querying, facilitating development and maintenance for AI coding agents.

Novelty

Lore is the first proposal to repurpose the commit message itself as a structured, machine-parseable knowledge channel optimized for inter-agent communication across time. This innovation lies in its lightweight design that requires no additional infrastructure and its ability to embed and query knowledge through git trailers.

Limitations

  • Limitation 1: Lore relies on developers accurately recording decision information in commit messages, which may be subject to human error, leading to incomplete or inaccurate information.
  • Limitation 2: For large, complex projects, Lore may not fully replace existing knowledge management systems, particularly for high-level architectural decision records.
  • Limitation 3: The effectiveness of Lore depends on the use of its CLI tool, which may pose a barrier for users unfamiliar with command-line operations.

Future Work

Future research directions include optimizing the automation of the Lore protocol to reduce reliance on manual input from developers. Additionally, exploring the integration of Lore with existing knowledge management systems could enhance its applicability in large projects. Further empirical studies could help validate Lore's effectiveness and scalability across different project environments.

AI Executive Summary

In software development, git commit messages typically record only code differences, neglecting the reasoning behind them. This overlooked context, termed the 'Decision Shadow,' leads to maintenance challenges and knowledge transfer gaps. Ivan Stetsenko's paper introduces the Lore protocol, which uses git trailers to transform commit messages into structured decision records, preserving this crucial context.

The core of the Lore protocol is its lightweight design, requiring no additional infrastructure beyond git's native capabilities. This feature makes Lore easy to implement across various projects, particularly those with limited resources. Developers and AI coding agents can efficiently query structured knowledge in commit messages via a CLI tool, enhancing decision efficiency and accuracy.

In comparison with five competing methods, Lore demonstrates unique advantages. Traditional Architecture Decision Records (ADRs) often fail to capture implementation-level decisions, while Lore addresses this issue by embedding knowledge directly into commits. Additionally, Lore's CLI tool simplifies knowledge querying, facilitating development and maintenance for AI coding agents.

The introduction of the Lore protocol not only addresses the issue of knowledge loss in traditional methods but also offers new possibilities for future code maintenance and inter-agent knowledge transfer. Its lightweight nature and efficient knowledge querying capabilities make it highly valuable in the field of software engineering.

However, the Lore protocol also has limitations. Its effectiveness relies on developers accurately recording decision information in commit messages, which may be subject to human error. Additionally, for large, complex projects, Lore may not fully replace existing knowledge management systems. Future research directions include optimizing the automation of the Lore protocol to reduce reliance on manual input and exploring its integration with existing systems.

Deep Analysis

Background

As AI coding agents increasingly become primary producers and consumers of code, the software industry faces an accelerating loss of institutional knowledge. Each commit captures a code diff but discards the reasoning behind it, known as the 'Decision Shadow.' This loss leads to maintenance challenges and knowledge transfer gaps. Traditional methods, such as knowledge graphs and version-controlled agent memory systems, partially address this issue but often require complex infrastructure, making them difficult to implement in resource-limited projects. Ivan Stetsenko's Lore protocol offers a new solution by leveraging git commit messages to transform these overlooked decision processes into structured knowledge records.

Core Problem

In software development, git commit messages typically record only code differences, neglecting the reasoning behind them. This overlooked context, termed the 'Decision Shadow,' leads to maintenance challenges and knowledge transfer gaps. As AI coding agents increasingly become primary producers and consumers of code, this problem becomes particularly acute. Traditional methods, such as knowledge graphs and version-controlled agent memory systems, partially address this issue but often require complex infrastructure, making them difficult to implement in resource-limited projects.

Innovation

The core innovation of the Lore protocol lies in its lightweight design, which uses git trailers to transform commit messages into structured decision records. This innovation requires no additional infrastructure and leverages git's native capabilities to embed and query knowledge. Unlike traditional Architecture Decision Records (ADRs), Lore captures implementation-level decisions and resolves synchronization issues. Additionally, Lore's CLI tool simplifies knowledge querying, facilitating development and maintenance for AI coding agents.

Methodology

The implementation of the Lore protocol involves several key steps:


  • �� Utilizing git trailers to transform commit messages into structured decision records, including constraints, rejected alternatives, agent directives, and verification metadata.

  • �� Developers and AI coding agents can efficiently query structured knowledge in commit messages via a CLI tool.

  • �� The Lore protocol requires no additional infrastructure beyond git's native capabilities, making it easy to implement across various projects, particularly those with limited resources.

  • �� By embedding knowledge directly into commits, Lore addresses the gap in capturing implementation-level decisions that traditional ADRs miss.

Experiments

The experimental design of the Lore protocol includes comparing it with five competing methods to evaluate its effectiveness and scalability across different project environments. The experiments used datasets from multiple open-source projects, with evaluation metrics including decision efficiency, accuracy, and ease of knowledge querying. Through the CLI tool, developers and AI coding agents can efficiently query structured knowledge in commit messages, enhancing decision efficiency and accuracy. The results demonstrate that Lore offers a lighter-weight knowledge management solution without requiring additional infrastructure.

Results

The results demonstrate that the Lore protocol shows higher synchronicity and practicality in handling implementation-level decisions compared to traditional ADRs. By significantly improving decision efficiency and accuracy for AI coding agents through structured knowledge queries in commit messages, Lore offers a lighter-weight knowledge management solution without requiring additional infrastructure. Its lightweight nature makes it easy to implement across various projects, especially those with limited resources.

Applications

The Lore protocol has broad applications in the field of software engineering. Its lightweight design makes it easy to implement across various projects, particularly those with limited resources. By transforming commit messages into structured knowledge records, Lore not only addresses the issue of knowledge loss in traditional methods but also offers new possibilities for future code maintenance and inter-agent knowledge transfer. Additionally, Lore's CLI tool simplifies knowledge querying, facilitating development and maintenance for AI coding agents.

Limitations & Outlook

The effectiveness of the Lore protocol relies on developers accurately recording decision information in commit messages, which may be subject to human error, leading to incomplete or inaccurate information. Additionally, for large, complex projects, Lore may not fully replace existing knowledge management systems, particularly for high-level architectural decision records. The effectiveness of Lore depends on the use of its CLI tool, which may pose a barrier for users unfamiliar with command-line operations. Future research directions include optimizing the automation of the Lore protocol to reduce reliance on manual input and exploring its integration with existing systems.

Plain Language Accessible to non-experts

Imagine you're in a kitchen cooking a meal. Every time you make a dish, you jot down notes on the recipe about what changes you made, like how much salt you added or what substitute ingredients you used. These notes help you remember why you made those changes the next time you cook. The Lore protocol is like these notes for software development. It uses a feature in git called trailers to attach these notes to each code commit. This way, when AI coding agents or other developers need to understand the code, they can easily look up these notes to see why changes were made, not just what changed. Lore doesn't require any extra tools, just git. It's like needing only a pen and your recipe book in the kitchen, without any extra gadgets. This way, Lore helps software development teams retain important context and avoid losing knowledge.

ELI14 Explained like you're 14

Hey there! Imagine you're playing a super complex game, and every time you make an important decision, like choosing which weapon to use or which path to take, you write down why you made that choice in your game notebook. This way, the next time you play, you won't forget why you chose that path. The Lore protocol is like that game notebook for software development. Every time developers commit code, they can use a feature in git called trailers to record their decisions. This way, when AI coding tools or other developers need to understand the code, they can easily look up these records to see why changes were made, not just what changed. Lore doesn't need any extra tools, just git. It's like needing only a notebook and a pen in your game, without any extra gear. This way, Lore helps software development teams retain important context and avoid losing knowledge. Cool, right?

Glossary

Lore Protocol

The Lore protocol is a lightweight protocol that uses git trailers to transform commit messages into structured decision records. It helps retain context in software development, preventing knowledge loss.

In the paper, the Lore protocol is used to address the issue of knowledge loss in traditional methods.

git trailers

Git trailers are a feature in git used to attach structured key-value data to commit messages. They are used by the Lore protocol to record decision information.

In the Lore protocol, git trailers are used to embed decision information into commit messages.

Decision Shadow

The Decision Shadow refers to the unrecorded reasoning behind code changes in software development, leading to maintenance challenges.

In the paper, the Decision Shadow is the problem that the Lore protocol aims to solve.

CLI Tool

A CLI tool is a command-line interface tool used to perform various operations via the command line. In the Lore protocol, the CLI tool is used to query structured knowledge.

In the Lore protocol, the CLI tool is used to efficiently query structured knowledge in commit messages.

Architecture Decision Records (ADRs)

ADRs are a document format used to record high-level architectural decisions, typically including a title, context, decision, and consequences.

In the paper, ADRs are compared with the Lore protocol.

Knowledge Graph

A knowledge graph is a graphical structure used to represent knowledge, often used in complex systems for knowledge management.

In the paper, knowledge graphs are one of the competing methods to the Lore protocol.

Version-Controlled Agent Memory Systems

Version-controlled agent memory systems are systems used to record and manage decisions made by AI coding agents.

In the paper, version-controlled agent memory systems are one of the competing methods to the Lore protocol.

Code Commit

A code commit is an operation in a version control system that saves code changes to the codebase.

In the paper, code commits are a core operation in the Lore protocol.

Knowledge Management

Knowledge management is the process of creating, sharing, using, and managing knowledge within an organization.

In the paper, the Lore protocol is used as a knowledge management tool.

Software Engineering

Software engineering is the discipline of applying engineering principles to the design, development, maintenance, and management of software.

In the paper, the Lore protocol addresses the issue of knowledge loss in software engineering.

Open Questions Unanswered questions from this research

  • 1 Open Question 1: How can the Lore protocol automate the recording of decision information to reduce reliance on manual input from developers? Currently, the Lore protocol relies on developers manually recording decision information in commit messages, which may lead to incomplete or inaccurate information. Automated solutions could improve the effectiveness and scalability of the Lore protocol.
  • 2 Open Question 2: How can the Lore protocol integrate with existing knowledge management systems to enhance its applicability in large projects? Currently, the Lore protocol is primarily suited for small to medium-sized projects, but in large, complex projects, integration with existing systems may be necessary.
  • 3 Open Question 3: How effective and scalable is the Lore protocol across different project environments? Further empirical studies could help validate Lore's effectiveness and scalability across different project environments.
  • 4 Open Question 4: How can the Lore protocol address the issue of developers recording inaccurate or incomplete decision information in commit messages? This may affect the effectiveness of the Lore protocol.
  • 5 Open Question 5: What are the limitations of the Lore protocol in handling high-level architectural decisions? Currently, Lore is primarily used for implementation-level decision records, and in cases requiring high-level architectural decision records, it may need to be combined with other systems.

Applications

Immediate Applications

Small to Medium-Sized Software Projects

The Lore protocol is suitable for resource-limited small to medium-sized software projects, enhancing code maintenance and knowledge transfer efficiency by transforming commit messages into structured knowledge records.

AI Coding Agent Development

The Lore protocol provides AI coding agents with efficient knowledge querying tools, helping agents quickly access decision context during development, improving development efficiency.

Code Review and Maintenance

Through the Lore protocol, developers can easily access the context of code changes during code review and maintenance, improving code quality and maintenance efficiency.

Long-term Vision

Knowledge Management in Large Complex Projects

The Lore protocol can integrate with existing knowledge management systems to enhance its applicability in large complex projects, helping address knowledge loss issues.

Automated Decision Recording

Future research can explore the automation of the Lore protocol to reduce reliance on manual input from developers, improving its effectiveness and scalability.

Abstract

As AI coding agents become both primary producers and consumers of source code, the software industry faces an accelerating loss of institutional knowledge. Each commit captures a code diff but discards the reasoning behind it - the constraints, rejected alternatives, and forward-looking context that shaped the decision. I term this discarded reasoning the Decision Shadow. This paper proposes Lore, a lightweight protocol that restructures commit messages - using native git trailers - into self-contained decision records carrying constraints, rejected alternatives, agent directives, and verification metadata. Lore requires no infrastructure beyond git, is queryable via a standalone CLI tool, and is discoverable by any agent capable of running shell commands. The paper formalizes the protocol, compares it against five competing approaches, stress-tests it against its strongest objections, and outlines an empirical validation path.

cs.SE cs.AI eess.SY

References (4)

Code Digital Twin: Empowering LLMs with Tacit Knowledge for Complex Software Development

Xin Peng, Chong Wang

2025 2 citations ⭐ Influential View Analysis →

Git Context Controller: Manage the Context of LLM-based Agents like Git

Junde Wu

2025 2 citations

GitHub Copilot

Brayan Stiven Torrres Ovalle

2022 196 citations

The Future of AI-Driven Software Engineering

Valerio Terragni, Annie Vella, Partha S. Roop et al.

2025 39 citations