Create new Agent Skills interactively or from templates. Use when user wants to create, generate, scaffold, or build a new skill, or mentions creating skills, writing skills, skill templates, skill development.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
The command stays on one line. Scroll horizontally to inspect it before copying.
Prefer a local copy? Download the files currently available to SkillsMP.
File Explorer
16 files
Showing SKILL.md
SKILL.md
Source instructions · Read-only preview
name
skill-creator
description
Create new Agent Skills interactively or from templates. Use when user wants to create, generate, scaffold, or build a new skill, or mentions creating skills, writing skills, skill templates, skill development.
allowed-tools
["Read","Write","Edit"]
Skill Creator
Creates new Agent Skills through interactive wizard or template-based generation. A meta-skill for bootstrapping Claude Code capabilities.
When This Skill Activates
User asks to "create a skill", "make a skill", "generate a skill", "scaffold a skill"
User mentions "skill template", "new skill", "custom skill", "skill development"
User wants to extend Claude with custom capabilities or workflows
User asks about building, authoring, or writing skills
Workflow
Step 1: Choose Mode
Ask the user which mode they prefer:
Interactive Wizard: Guided Q&A to build a skill from scratch
Template-Based: Start from a pre-built template and customize
Hybrid: Template foundation + interactive customization
Step 2: Gather Requirements (Interactive Mode)
Ask these questions in order:
Skill Name: What should the skill be called? (kebab-case, max 64 chars)
Purpose: What does this skill do? (one sentence)
Trigger Phrases: When should Claude use this skill? (list 3-5 specific phrases users might say)
Tool Permissions: Which tools does the skill need? (Read, Write, Edit, Bash, Grep, Glob - request minimal)
Skill Type:
Simple: Single SKILL.md only
With Scripts: SKILL.md + Python/bash helper scripts
With Reference: SKILL.md + reference.md for technical docs
With Examples: SKILL.md + examples.md for use cases
Processes data files (CSV, JSON, YAML). Validates schemas, transforms data, handles large files, generates outputs.
Best Practices for Generated Skills
Description Writing
Start with "Use when..." for clarity
Include 3-5 specific trigger scenarios
Mention what makes this skill unique
Keep under 1024 characters
Tool Permissions
Request only necessary tools
Use specific constraints (e.g., Bash(python:*))
Explain why each tool is needed
Progressive Disclosure
Keep SKILL.md focused (under 500 lines)
Reference supporting files for deep details
Don't duplicate content across files
Common Patterns
Skill with Scripts
When a skill needs helper scripts:
## Utility Scripts
This skill includes helper scripts for complex operations:
```bash
# Validate input
python scripts/validate.py input.txt
# Transform data
python scripts/transform.py input.csv output.json
### Skill with Reference
For technical/API skills, reference detailed docs:
```markdown
## Technical Reference
For complete API details, see [reference.md](reference.md).
Skill with Environment Variables
If the skill needs environment setup:
## Requirements
Set these environment variables:
-`API_KEY`: Your service API key
-`ENDPOINT`: API endpoint URL
Install dependencies:
```bash
pip install requests pydantic
## Troubleshooting
### Skill Not Appearing
- Verify SKILL.md path is correct
- Check YAML frontmatter is valid
- Run `claude --debug` to see loading errors
### Description Not Triggering
- Add more specific trigger phrases
- Include keywords users would naturally say
- Test with different query phrasings
### Validation Errors
- Check name is kebab-case only
- Verify description length < 1024 chars
- Ensure YAML has proper `---` delimiters
## Examples
See [examples.md](examples.md) for complete walkthroughs of creating skills in each mode.
