Create a new custom skill from scratch — research the practice, then write a structured skill with prerequisites, process and output that follows singularskills conventions. Use when the user repeats a task no existing skill covers, in any domain, and wants a reusable skill written for it. Not for finding an existing one; see /singular-find-skill.
Installs into .claude/skills of the current project.
Are you the author of Singular Create Skill?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/bromso-singular-create-skill)
---
name: singular-create-skill
description: Create a new custom skill from scratch — research the practice, then write a structured skill with prerequisites, process and output that follows singularskills conventions. Use when the user repeats a task no existing skill covers, in any domain, and wants a reusable skill written for it. Not for finding an existing one; see /singular-find-skill.
metadata:
argument-hint: "<topic>"
disable-model-invocation: true
---
# Create Skill
Create a new custom skill for "$ARGUMENTS" following the singularskills SKILL.md pattern. Research best practices and generate a structured skill.
## Prerequisites
None — utility skill, run anytime.
## Process
1. **Understand the skill:**
- What does "$ARGUMENTS" need to accomplish?
- Ask the user to clarify if ambiguous: Who uses it? What's the input? What's the output?
2. **Research best practices:**
- Use WebSearch to find current best practices, patterns, and conventions for "$ARGUMENTS"
- Look for: official documentation, community guides, expert recommendations
- Identify the key steps a practitioner would follow
3. **Create the SKILL.md:**
- **Frontmatter:** `name` (the skill's directory name); `description` shaped as `<Verb-first sentence saying what it does>. Use when <the user's situation, at least six words>.` optionally followed by `Not for <adjacent thing>; see /<domain>-<skill>.`, 120–400 characters; `argument-hint` when the skill takes an argument
- **Title:** Clear skill name as H1
- **Introduction:** One sentence explaining what the skill does with "$ARGUMENTS" placeholder
- **Prerequisites:** What files or context are needed (or "None")
- **Process:** Numbered steps with clear instructions
- What to read
- What to research (WebSearch if applicable)
- What to produce
- What to verify
- **Output:** Where the artifact is written and what to present to the user
4. **Validate the skill:**
- `name` matches the directory name
- Description starts with a recognized verb, has a "Use when" clause of at least six words, and is 120–400 characters
- `argument-hint` is present if the body uses "$ARGUMENTS"
- Body under 2000 words
- Artifact path uses `.singularskills/` convention
5. **Install the skill:**
- Write to `.claude/skills/<skill-name>/SKILL.md`
- Report the skill name and invocation command
## Output
The new skill written to `.claude/skills/<skill-name>/SKILL.md`. Present the skill summary and how to invoke it. The skill is available as `/<name>` in the next Claude Code session.