Skip to content

tmdgusya/lora

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

Lore

Git commit messages as a structured knowledge protocol for AI coding agents.

Based on the paper "Lore: Repurposing Git Commit Messages as a Structured Knowledge Protocol for AI Coding Agents" by Ivan Stetsenko (2026).

Problem: Decision Shadow

Every commit destroys decision context. Developers and AI agents consider alternatives, evaluate tradeoffs, and choose one approach — but only the final diff survives. The constraints, rejected alternatives, confidence levels, and warnings to future modifiers all evaporate.

This lost context is the Decision Shadow. Individually small, but accumulated it creates what the industry calls "legacy code" — code that works but whose structural rationale has been lost.

Solution: Lore Atoms

Lore turns each commit into an atomic unit of institutional knowledge by encoding decision context as git trailers — a native git feature that requires no custom tooling.

Prevent silent session drops during long-running operations

The auth service returns inconsistent status codes on token
expiry, so the interceptor catches all 4xx responses and
triggers an inline refresh.

Constraint: Auth service does not support token introspection
Constraint: Must not add latency to non-expired-token paths
Rejected: Extend token TTL to 24h | security policy violation
Rejected: Background refresh on timer | race condition
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Error handling is intentionally broad (all 4xx)
  -- do not narrow without verifying upstream behavior
Tested: Single expired token refresh (unit)
Not-tested: Auth service cold-start > 500ms behavior

Trailer Vocabulary

Trailer Purpose
Constraint: External limit that shaped the decision
Rejected: Alternative considered and why (alt | reason)
Confidence: high / medium / low
Scope-risk: narrow / moderate / broad
Reversibility: clean / moderate / difficult
Directive: Warning or instruction for future modifiers
Tested: What was verified
Not-tested: Known coverage gaps
Related: Linked commits forming a decision chain

All trailers are optional and repeatable. Extend with custom trailers as needed.

Querying

Git trailers are queryable with standard git tools:

# Find commits containing a specific trailer
git log --all --grep="^Constraint:" -- path/to/file.ts
git log --all --grep="^Rejected:" -- path/to/file.ts
git log --all --grep="^Directive:" -- path/to/file.ts

Zero infrastructure. Your git history becomes a queryable decision database.

Install

npx skills add tmdgusya/lora

This installs a Claude Code skill that teaches AI agents to write Lore-formatted commit messages. Works with any agent that supports the skills ecosystem.

Global install (available across all projects):

npx skills add tmdgusya/lora -g

Key Properties

  • Atomic binding — knowledge fused permanently with code changes
  • Temporal immutability — append-only log, once committed it persists forever
  • Universal availability — every git project already has this channel
  • Natural granularity — commits are already scoped to logical work units
  • Graceful degradation — worst case is identical to current practice

License

MIT

About

Lore: Git 커밋 메시지를 AI 코딩 에이전트용 구조화된 지식 프로토콜로 변환합니다. 제약조건, 거절된 대안, 신뢰도, 지시사항을 git trailers로 인코딩하여 git 히스토리를 쿼리 가능한 의사결정 데이터베이스로 만듭니다. | Structured knowledge protocol that turns git commits into queryable decision databases using native git trailers.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages