How to use the Skill Creation Assistant agent skill

What a skill is and what this skill makes the agent do

A skill is a reusable playbook: step-by-step instructions the chat agent reads on demand with its read_skill tool. The agent only sees each skill's name and description in its system prompt, so the description acts as the trigger that tells it when the skill applies.

When this skill is active, the agent asks at most one short question if a key detail is missing. Otherwise it infers sensible defaults and drafts a skill with a clear name, a trigger description, step-by-step instructions, and optional reference files. It calls propose_creation to show the draft as a card. Nothing is saved automatically; you edit or confirm the card.

• Drafts a name and a trigger description
• Writes instructions as ordered steps with concrete examples and edge cases
• Moves large reference material into references/*.md files
• Shows the draft as a preview card for you to edit or confirm

Install the skill from the ZIP or the Claude apps

Download the ZIP from this page and unzip it into one of two locations. Use ~/.claude/skills/ to make it available across your projects, or a project's .claude/skills/ folder to limit it to that project.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/
• Project only: unzip into .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Requirements before you start

This skill needs the DeepTutor runtime with its chat agent, plus the read_skill and propose_creation tools. It does not run scripts and does not use the network. It works from your plain-language description and any materials you share in the conversation, using the agent's other tools such as web search, knowledge base, or attachments when you provide them.

• DeepTutor runtime with its chat agent
• read_skill and propose_creation tools available
• Runs scripts: no
• Uses network: no

How to trigger it with natural requests

Ask the agent in plain language to create, write, or draft a skill for a workflow you repeat. Describe the task, who it is for, and what a good result looks like. You can include examples or attach material for the agent to use.

The agent writes the draft in the same language you use. Once the card appears, ask for changes in ordinary words and the agent revises the draft under the same name.

• "Make a skill that turns my lecture notes into a five-question quiz."
• "Create a reusable skill for checking citations in my essays against a reference list."
• "Draft a skill for summarising a chapter in the format I use for study guides."

Skills versus Personas: what this skill is not for

This skill is for multi-step workflows. If you want to set a tone or personality, such as always answering like a strict tutor, that is a Persona, not a Skill. The skill file tells the agent to explain this briefly and not force the request into a skill.

Because the drafting step only produces a preview card, a skill is never saved automatically. If you need one saved without review, this skill is not the right fit.

• Use a Skill for repeatable procedures and task playbooks
• Use a Persona for tone, style, and behaviour settings

Safety notes: scripts and network access

According to the listing, this skill runs no scripts and uses no network access. Its output is a text draft shown to you on a card. Review the name, description, and instructions before you confirm, since the draft becomes a skill only after you accept it.

Check any reference files the agent proposes as well, because they become part of the skill's content once saved.

• No scripts are run by this skill
• No network access is used by this skill
• Review every draft card before confirming