Skill Creator

Skill Creator

Create reusable Skills for the agent.

Understand Skills

A Skill is a self-contained directory of instructions that teaches the agent how to perform a particular kind of task. Skills use progressive disclosure:

1. The agent first sees only the Skill name, description, and path.
2. When a task matches, the agent reads the full SKILL.md.
3. Supporting files are accessed later only when needed.

A Skill provides knowledge and workflow instructions. It does not grant new permissions, bypass safeguards, or require a particular tool.

Create the required structure

Create each Vault Skill at:

.agents/skills/<skill-name>/SKILL.md

The directory name and frontmatter name must match exactly. Every Skill requires YAML frontmatter followed by Markdown instructions:

---
name: example-skill
description: Explain what this Skill does and when the agent should use it.
---

# Example Skill

Follow the workflow described here.

Choose the name

Choose a short, action-oriented name that:

• Contains only lowercase ASCII letters, digits, and single hyphens.
• Starts and ends with a letter or digit.
• Is no longer than 64 characters.
• Exactly matches its directory name.

Write the description

Treat description as the triggering contract. State both what the Skill does and when it should be used. Keep it specific enough to distinguish the Skill from unrelated tasks. The description must be non-empty and no longer than 1024 characters.

Do not put essential triggering conditions only in the body because the body is not loaded until after the Skill has been selected.

Write effective instructions

Assume the agent is already capable. Include only specialized, procedural, project-specific, or easy-to-miss information. Write direct, actionable instructions and prefer concise workflows or examples over broad explanations.

Match detail to the task: give flexible guidance when several approaches are valid, ordered steps when sequence matters, and exact commands or scripts only when deterministic behavior is required.

Do not unnecessarily restrict which tools the agent may use. Let it choose from the tools available in its environment unless a particular integration is an essential part of the Skill.

Add optional resources only when useful

skill-name/
├── SKILL.md
├── scripts/
├── references/
└── assets/

• Use scripts/ for repeatable deterministic operations.
• Use references/ for detailed knowledge loaded only when needed.
• Use assets/ for templates and files used in generated output.

Supporting files are not automatically loaded. Reference them clearly from SKILL.md and explain when to read or use them. Avoid unrelated files such as README files, changelogs, or installation guides.

Follow the creation workflow

1. Determine the intended task and representative user requests.
2. Choose and validate the Skill name.
3. Check whether the target directory already exists.
4. Plan the minimum required instructions and supporting resources.
5. Create the directory and SKILL.md with any suitable available tools.
6. Add optional resources only when they provide reusable value.
7. Read the resulting files and verify their structure and content.
8. Report the created path and briefly describe when the Skill will trigger.

Do not overwrite an existing Skill without making the conflict clear to the user.

Validate the result

Verify that:

• The file is at .agents/skills/<name>/SKILL.md.
• The directory and frontmatter names match.
• name and description are present and valid.
• The body contains actionable instructions.
• Every referenced supporting file exists.
• SKILL.md is UTF-8 text, contains no null bytes, and is smaller than 64 KiB.