moe-discipline
Use during any implementation task to shape HOW you work — minimal changes, proportional responses, scope containment, and honest gap reporting; fires alongside the task itself rather than at completion time
- License 许可
- Apache-2.0
- 所属仓库
- zak-keown/moe
- 最近更新
- 2026-09-05
- Stars
- ★ 0
- Forks
- 1
01 前言 Front Matter
- name
- moe-discipline
- description
- Use during any implementation task to shape HOW you work — minimal changes, proportional responses, scope containment, and honest gap reporting; fires alongside the task itself rather than at completion time
03 SKILL.md 正文
Moe Discipline
Overview
Core principle: do exactly what was asked, nothing more, and say what you did not do.
verify-completion fires when you are about to claim the work
is done. This skill fires while you are doing it. The two are complements:
one shapes the work, the other gates the claim.
The Seven Rules
-
Smallest-change bias. The correct change is the smallest one that solves the stated problem. A refactor is not a fix; a rewrite is not a refactor. If the task says "fix the null check", touch the null check.
-
Edit over create. Before writing a new file, read what already exists. An edit to an existing file is almost always better than a new file that duplicates, shadows, or contradicts it.
-
Read before writing. No file is created or edited without first reading the files it will live beside. No function is added without grepping for one that already does the job.
-
Match your response to the task. A one-line fix gets a one-line answer. A rename gets a rename, not a paragraph explaining why naming matters. The length of the reply tracks the complexity of the work, not the importance you assign to it.
-
Scope stays where it started. If the task does not mention tests, do not write tests. If it does not mention docs, do not update docs. If you see something unrelated that needs fixing, name it; do not fix it inside this task.
-
When uncertain, do less. A change you are unsure about is a change the reviewer has to debug. If you cannot tell whether a line should change, leave it alone and say why you left it.
-
Name what you did not do. Every task has edges you chose not to cross. Say which edges and why. A gap you name is an informed decision; a gap you hide is a latent defect.
Red Flags
| Signal | What is happening |
|---|---|
| A diff that touches files the task did not name | Scope has escaped |
| A new file where an existing one could have been edited | Create-over-edit bias |
| An answer longer than the change it describes | Proportionality lost |
| A "while I was here" aside that lands as code | Gold-plating |
| A refactor folded into a bug fix | Two changes in one, neither reviewable |
| Silence about what was left unchanged | Gap hiding |
Rationalization Prevention
| Excuse | Reality |
|---|---|
| "It was easy so I just did it" | Easy changes still carry review cost |
| "They will need this eventually" | Eventually is not now |
| "It was already broken" | A separate finding, not this task |
| "The code was right there" | Proximity is not a mandate |
| "I cleaned it up while I was in the file" | A cleanup is a separate change |
| "It is only a few lines" | Line count does not determine scope |
| "The tests needed updating too" | Only if the task said so |
04 同仓库更多 Skills
write-skill
Use when creating new skills, editing existing skills, or verifying skills work before deployment
write-plan
Use when you have a spec or requirements for a multi-step task, before touching code
write-clearly
Apply Strunk's timeless writing rules to ANY prose humans will read—documentation, commit messages, error messages, explanations, reports, or UI text. Makes your writing clearer, stronger, and more professional.
windows-vm
Create, manage, or connect to a headless Windows 11 VM running in Docker with SSH access. Use when the user wants to spin up, stop, restart, or SSH into a Windows VM.
verify-completion
Use when about to claim work is complete, fixed, or passing, before committing or creating MRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always
using-moe
Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions