How do I install the Scope skill?

Run npx skillmds@latest add mithraeums/scope in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Scope safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Scope?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Scope free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

mithraeums (@mithraeums) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Scope by mithraeums — Agent Skill

Name: Scope
Author: mithraeums

Scope

Turn a vague request into a scoped plan before building — explicit goal, explicit non-goals, the smallest viable first step. Use at the start of any non-trivial task, when a request is ambiguous or could balloon, or when you catch yourself about to over-build. Stops scope creep and wasted work. Host-agnostic. Trigger: /scope, "plan this", "before we build", "what's the smallest version", "is this in scope".

mithraeums 9922696 2.2 KB Updated Sep 17, 2026

File contents

scope — plan before you build

What this is

A short discipline that runs before implementation: decide what you're building, what you're explicitly not building, and the smallest step that proves the direction. Five minutes here saves an hour of building the wrong thing.

When to use

Any non-trivial or multi-step task.
The request is vague, broad, or could expand ("make it flexible", "handle everything").
You feel the urge to add abstraction, config, or "while I'm here" extras.

The pass

Restate the goal in one sentence — the user-visible outcome, not the mechanism. Confirm if unsure.
List non-goals — explicitly name what you will NOT do this round. This is the highest-value step; it's where creep dies.
Smallest viable step — the least work that delivers or de-risks the goal. Ship/learn from it before expanding.
Surface unknowns — the 1-3 questions whose answers change the plan. Ask only those; don't interrogate.
Name the verify — how you'll know it worked (run it, a test, an observed behavior).

Output

Goal: <one sentence, user-visible>
Non-goals: <what we're deliberately not doing now>
First step: <smallest thing that delivers/de-risks>
Open questions: <only the ones that change the plan>
Verify: <how we'll know it worked>

Default to less

No abstraction for single-use code.
No config/flags nobody asked for.
No rewrite disguised as a refactor.
"Flexibility" and "future-proofing" are non-goals unless requested.
One working slice beats a half-built general system.

Notes

Pure planning protocol; any agent. The standalone essence of corp's CEO. The Verify line is the contract you hold yourself to before claiming the goal is met — don't declare done until you've run it.

---name: scopedescription: Turn a vague request into a scoped plan before building — explicit goal, explicit non-goals, the smallest viable first step. Use at the start of any non-trivial task, when a request is ambiguous or could balloon, or when you catch yourself about to over-build. Stops scope creep and wasted work. Host-agnostic. Trigger: /scope, "plan this", "before we build", "what's the smallest version", "is this in scope".---# scope — plan before you build## What this isA short discipline that runs *before* implementation: decide what you're building,what you're explicitly **not** building, and the smallest step that proves thedirection. Five minutes here saves an hour of building the wrong thing.## When to use- Any non-trivial or multi-step task.- The request is vague, broad, or could expand ("make it flexible", "handle everything").- You feel the urge to add abstraction, config, or "while I'm here" extras.## The pass1. **Restate the goal** in one sentence — the user-visible outcome, not the mechanism. Confirm if unsure.2. **List non-goals** — explicitly name what you will NOT do this round. This is the highest-value step; it's where creep dies.3. **Smallest viable step** — the least work that delivers or de-risks the goal. Ship/learn from it before expanding.4. **Surface unknowns** — the 1-3 questions whose answers change the plan. Ask only those; don't interrogate.5. **Name the verify** — how you'll know it worked (run it, a test, an observed behavior).## Output```Goal: <one sentence, user-visible>Non-goals: <what we're deliberately not doing now>First step: <smallest thing that delivers/de-risks>Open questions: <only the ones that change the plan>Verify: <how we'll know it worked>```## Default to less- No abstraction for single-use code.- No config/flags nobody asked for.- No rewrite disguised as a refactor.- "Flexibility" and "future-proofing" are non-goals unless requested.- One working slice beats a half-built general system.## NotesPure planning protocol; any agent. The standalone essence of `corp`'s CEO. The`Verify` line is the contract you hold yourself to before claiming the goal is met— don't declare done until you've run it.

Frequently asked questions