tmdgusya/lora/lore-commits
Claude Code and agent skill by tmdgusya/lora
- community
Checking that the SKILL.md file is still in the repository.
Use when writing git commit messages for non-trivial changes — captures decision context (constraints, rejected alternatives, confidence, directives) as structured git trailers so future agents and developers can query project knowledge via git log --trailer=
Install
npx vskill@latest install tmdgusya/lora/lore-commitsSKILL.md
Lore Commits
Overview
Every commit destroys decision context. You considered alternatives, evaluated tradeoffs, and chose one approach — but only the final diff survives. Lore reclaims this lost knowledge by encoding it as git trailers in commit messages.
Core principle: A commit is not a label for a diff — it's an atomic unit of institutional knowledge. Use git's native trailer format so decisions are machine-queryable, not buried in prose.
When to Use
- Any commit where you made non-obvious decisions
- When you rejected alternative approaches
- When external constraints shaped your implementation
- When future modifiers need warnings about fragile assumptions
- When test coverage has known gaps
When NOT to use: Trivial changes (typo fixes, formatting) where no real decisions were made.
The Format
All trailers are optional. Include only those that carry signal for this specific commit.
Trailer Vocabulary
| Trailer | Purpose | Example | |||
|---|---|---|---|---|---|
Constraint: | External limit that shaped the decision | Auth service does not support token introspection | |||
Rejected: | Alternative considered and why it was dropped. Use ` | ` to separate alternative from reason | `Extend token TTL to 24h | security policy violation` | |
Confidence: | How sure you are this is right: high, medium, low | high | |||
Scope-risk: | Blast radius: narrow, moderate, broad | narrow | |||
Reversibility: | How easy to undo: clean, moderate, difficult | clean | |||
Directive: | Warning to future modifiers — what NOT to do or what to check first | Do not narrow 4xx handling without verifying upstream behavior | |||
Tested: | What was verified and how | Single expired token refresh (unit) | |||
Not-tested: | Known gaps in verification | Auth service cold-start > 500ms behavior | |||
Related: | Linked commits forming a decision chain | abc1234 (initial auth interceptor) |
You can repeat trailers (multiple Constraint:, multiple Rejected:, etc.).
You can extend the vocabulary with custom trailers when needed (e.g., Deadline:, Stakeholder:, Migration:).
Complete Example
Why Git Trailers (Not Prose)
Agents naturally capture decision context as prose paragraphs. The problem: prose is not queryable.
Git trailers are a native git feature. Queryable with standard git tools:
This turns your git history into a queryable decision database at zero infrastructure cost.
Agent Workflow
When modifying code, before committing:
- Harvest context — What constraints did you face? What alternatives did you reject?
- Write the summary line — Imperative, focused on why not what
- Add body if the how needs explanation
- Append trailers — One per decision fact. Only include trailers that carry signal.
- Self-check — Would a future agent modifying this code benefit from knowing this?
Common Mistakes
| Mistake | Fix | ||
|---|---|---|---|
| Writing trailers as prose paragraphs | Use Key: value format on separate lines | ||
| Putting trailers before the body | Trailers must come after the body, separated by a blank line | ||
Omitting the reason in Rejected: | Always use `alternative | reason` format | |
| Adding trailers to trivial commits | Only add when real decisions were made | ||
| Duplicating diff content in trailers | Trailers capture what's NOT in the diff — the why, not the what | ||
Using Confidence: yes | Use the scale: high, medium, low |
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/lore-commits/SKILL.md
Popularity
- 7-day trend score
- 0
Refreshed 18 Mar 2026, 7 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/tmdgusya/lora/lore-commits)