How do I install the Create Skill skill?

Run npx skillmds@latest add jonhill90/create-skill 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 Create Skill 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 Create Skill?

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 Create Skill free to use?

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

Who published Create Skill?

jonhill90 (@jonhill90) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Create Skill by jonhill90 — Agent Skill

Name: Create Skill
Author: jonhill90

Create Skill

Design, create, and validate portable Agent Skills with effective triggers, concise instructions, and reusable scripts, references, or assets. Use when adding a new skill, substantially restructuring an existing skill, or validating a skill's structure, frontmatter, links, and bundled scripts.

jonhill90 660379e 2 files · 4.6 KB Updated Sep 17, 2026

File contents

Create a Skill

Create focused skills that another agent can discover and apply without loading unrelated context.

Ask where it belongs first

Before writing anything, settle where the skill will live. Most skills do not belong in a personal harness repository, and the answer changes what the skill has to earn.

Situation	Where	Has to earn its place?
Every project, every day	personal roster	yes — full evidence bar
One repository only	that repo's `.claude/skills/` or `.agents/skills/`	no
Occasional domain work	nothing installed — `npx skills use <pkg>@<skill>`	no

A skill in the personal roster is loaded on every request on every harness, so it is the only case that needs a failing scenario behind it. A project skill costs nothing anywhere else — write it, commit it next to the code, and move on.

If the answer is "one repository" or "someone else already wrote it", stop here. The rest of this skill is about authoring for the portable case.

Workflow

Gather concrete examples of requests that should trigger the skill.
Define the repeated workflow and identify deterministic operations.
Create skills/<name>/SKILL.md.
Add only the resources needed by that workflow.
Validate the skill and test its scripts.
Exercise it against realistic trigger and non-trigger prompts.

Structure

skills/<name>/
├── SKILL.md
├── scripts/       Optional deterministic helpers
├── references/    Optional detail loaded on demand
└── assets/        Optional output resources

Do not add auxiliary files such as a skill-level README, changelog, or installation guide.

Frontmatter

Use portable frontmatter by default:

---
name: review-api
description: Review an HTTP API for correctness, compatibility, and security. Use when evaluating route, schema, authentication, or versioning changes.
---

Requirements:

Use lowercase letters, digits, and hyphens.
Keep the name at 64 characters or fewer.
Match the directory name exactly.
Describe both the capability and its trigger conditions.
Avoid harness-specific fields unless the skill explicitly documents reduced portability.

Body

Use imperative instructions.
Assume the agent already knows general software engineering.
Include procedural or domain knowledge that changes execution.
Keep the main file under 500 lines.
Put variant-specific detail in directly linked references.
Explain when each reference should be read.

Resource Selection

Use scripts/ when reliability requires a repeated deterministic operation. Test every script that is added or changed.

Use references/ for schemas, detailed procedures, provider-specific guidance, or material that is only relevant to some requests.

Use assets/ for templates and files used in produced output rather than loaded as instructions.

Validation

Run:

# inside this repository:
python3 scripts/validate_repository.py skills/<name>
# anywhere else — the portable check:
npx skills add . --list
python3 -m unittest discover -s tests -v

Then test at least:

a request that should activate the skill;
a nearby request that should not activate it;
one realistic task using any bundled script or reference.

---name: create-skilldescription: Design, create, and validate portable Agent Skills with effective triggers, concise instructions, and reusable scripts, references, or assets. Use when adding a new skill, substantially restructuring an existing skill, or validating a skill's structure, frontmatter, links, and bundled scripts.---# Create a SkillCreate focused skills that another agent can discover and apply without loadingunrelated context.## Ask where it belongs firstBefore writing anything, settle where the skill will live. Most skills donot belong in a personal harness repository, and the answer changes whatthe skill has to earn.| Situation | Where | Has to earn its place? ||---|---|---|| Every project, every day | personal roster | yes — full evidence bar || One repository only | that repo's `.claude/skills/` or `.agents/skills/` | no || Occasional domain work | nothing installed — `npx skills use <pkg>@<skill>` | no |A skill in the personal roster is loaded on every request on every harness,so it is the only case that needs a failing scenario behind it. A projectskill costs nothing anywhere else — write it, commit it next to the code,and move on.If the answer is "one repository" or "someone else already wrote it", stophere. The rest of this skill is about authoring for the portable case.## Workflow1. Gather concrete examples of requests that should trigger the skill.2. Define the repeated workflow and identify deterministic operations.3. Create `skills/<name>/SKILL.md`.4. Add only the resources needed by that workflow.5. Validate the skill and test its scripts.6. Exercise it against realistic trigger and non-trigger prompts.## Structure```textskills/<name>/├── SKILL.md├── scripts/ Optional deterministic helpers├── references/ Optional detail loaded on demand└── assets/ Optional output resources```Do not add auxiliary files such as a skill-level README, changelog, orinstallation guide.## FrontmatterUse portable frontmatter by default:```yaml---name: review-apidescription: Review an HTTP API for correctness, compatibility, and security. Use when evaluating route, schema, authentication, or versioning changes.---```Requirements:- Use lowercase letters, digits, and hyphens.- Keep the name at 64 characters or fewer.- Match the directory name exactly.- Describe both the capability and its trigger conditions.- Avoid harness-specific fields unless the skill explicitly documents reduced portability.## Body- Use imperative instructions.- Assume the agent already knows general software engineering.- Include procedural or domain knowledge that changes execution.- Keep the main file under 500 lines.- Put variant-specific detail in directly linked references.- Explain when each reference should be read.## Resource SelectionUse `scripts/` when reliability requires a repeated deterministic operation.Test every script that is added or changed.Use `references/` for schemas, detailed procedures, provider-specific guidance,or material that is only relevant to some requests.Use `assets/` for templates and files used in produced output rather than loadedas instructions.## ValidationRun:```bash# inside this repository:python3 scripts/validate_repository.py skills/<name># anywhere else — the portable check:npx skills add . --listpython3 -m unittest discover -s tests -v```Then test at least:- a request that should activate the skill;- a nearby request that should not activate it;- one realistic task using any bundled script or reference.

Frequently asked questions