Skip to content
Back to skills

Make Skill

ASecurity

Turns a repeating workflow into a new Minga skill or agent — interviews the user, writes SKILL.md or an agent file in the right format, tests it on a real example, and validates the repository. Use when the user says "сделай свой skill", "добавь агента", "хочу автоматизировать это", "make a skill", "add an agent".

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 13, 2026
ai-agentsgobashnodegit

Works with

  • cli

Security analysis

A100/100

Scanned October 13, 2026

npx -y skills add dmminga1983/minga --skill make-skill --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Make Skill?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Make Skill
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/dmminga1983-make-skill/badge)](https://www.skillsdirectory.com/skills/dmminga1983-make-skill)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: make-skill
description: Turns a repeating workflow into a new Minga skill or agent — interviews the user, writes SKILL.md or an agent file in the right format, tests it on a real example, and validates the repository. Use when the user says "сделай свой skill", "добавь агента", "хочу автоматизировать это", "make a skill", "add an agent".
argument-hint: "[workflow to capture]"
---

# Make a skill or agent / Новый skill или агент

Workflow: $ARGUMENTS

Reply in the user's language.

## Skill or agent?

- **Skill** (`skills/<name>/SKILL.md`) — a procedure: steps, checklists, output format. Runs in the main conversation; the user can call it as `/minga:<name>`.
- **Agent** (`agents/<name>.md`) — a role with its own context window and tool list, for focused or parallel work (review, research, verification).

## Steps

1. **Interview:** what triggers the workflow, inputs, steps, output format, common mistakes, what "good" looks like. Ask for one real example.
2. **Name:** lowercase letters, digits and hyphens, ≤ 64 characters, unique in the repo.
3. **Description** (the most important line — it decides when the skill is used): what it does + when to use it + trigger phrases in **both Russian and English**. Keep it under 1,000 characters.
4. **Body:** short imperative steps, the output format, rules and boundaries. Under 500 lines; move long checklists to separate files next to `SKILL.md` and link them.
5. **Agent frontmatter:** `name`, `description`, `tools` (least privilege — no `Write`/`Bash` for reviewers), `model` (`haiku` for simple, `sonnet` default, `opus` for hard reasoning), `color`.
6. **Test** on the real example; fix what went wrong.
7. **Validate:** `node tools/validate.mjs` must pass. Add the skill to the tables in `README.md` and `README.en.md` and to `skills/start/SKILL.md`.

Never put secrets, personal data or confidential client material into a skill — skills in a public repository are public.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.