# skills-maker

> Analyze completed tool-usage workflows from chat history and package them into reusable Agent Skills. Rejects development tasks and guides users through skill creation with interactive refinement.

- Author: zhing2006
- Repository: zhing2006/skills-maker
- Version: 20260127183133
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/zhing2006/skills-maker
- Web: https://mule.run/skillshub/@@zhing2006/skills-maker~skills-maker:20260127183133

---

---
name: skills-maker
description: Analyze completed tool-usage workflows from chat history and package them into reusable Agent Skills. Rejects development tasks and guides users through skill creation with interactive refinement.
license: Apache-2.0
compatibility: Works with Claude Code, GitHub Copilot, Cursor, and other Agent Skills compatible tools
metadata:
  author: zhing2006
  version: "0.1.0"
  repository: https://github.com/zhing2006/skills-maker
---

# Skills Maker

A meta-skill that transforms your successful workflows into reusable Agent Skills.

## Task Type Detection

Before creating a skill, analyze the conversation history to determine the task type.

### Development Tasks (REJECT)

Do NOT create skills for these task types:

- **Code Creation**: Writing new functions, classes, modules, or scripts from scratch
- **Application Building**: Developing new applications, libraries, or frameworks
- **Feature Implementation**: Adding new features or functionality to existing code
- **Algorithm Design**: Creating new algorithms or data structures

**Rejection Response:**

> I cannot create a skill from this conversation because it involves **development work** (creating new code/tools). Skills are designed to capture **repeatable workflows using existing tools**, not the process of building new ones.
>
> **Suggestion:** If you've created a useful tool, you can ask me to create a skill for *using* that tool in future workflows.

### Tool-Usage Tasks (ACCEPT)

Proceed with skill creation for these task types:

- **Command Execution**: Using CLI tools, scripts, or system commands
- **Data Processing**: Transforming, analyzing, or manipulating data with existing tools
- **Configuration Tasks**: Setting up environments, configuring systems
- **Repeatable Workflows**: Multi-step procedures using established methods
- **File Operations**: Organizing, converting, or batch processing files

## User Interaction Flow

When a valid tool-usage task is detected, use AskQuestion to gather requirements:

### Questions to Ask

1. **Workflow Summary**
   - "I identified the following workflow: [summary]. Is this accurate?"

2. **Skill Naming**
   - Suggest a name based on the workflow (lowercase, hyphens only)
   - "Suggested name: `{{suggested-name}}`. Would you like to use this or provide a different name?"

3. **Language Preference**
   - "Which language should I use for the skill documentation?"
   - Options: English (EN) / Chinese (ZH)

4. **Scope Refinement**
   - "Are there any steps you want to exclude or modify?"
   - "Should I include any additional context or notes?"

5. **Tool Inclusion**
   - If scripts/tools were used: "The following tools were used: [list]. Should I include them in the skill's scripts folder?"

## Skill Generation

### Output Location

Generate the new skill in the **parent directory** of skills-maker:

- If skills-maker is at: `.claude/skills/skills-maker/`
- New skill goes to: `.claude/skills/{{skill-name}}/`

### Directory Structure

Create the following structure for the new skill:

```txt
{{skill-name}}/
├── SKILL.md              # Main instructions
├── scripts/              # (if tools were used)
│   └── {{copied tools}}
└── references/           # (if extensive documentation needed)
    └── {{reference files}}
```

### Generation Process

1. **Read Reference Documents (REQUIRED)**
   - Read [references/SPECIFICATION.md](references/SPECIFICATION.md) to understand the complete Agent Skills format
   - Read [references/BEST-PRACTICES.md](references/BEST-PRACTICES.md) to follow authoring guidelines
   - These documents contain critical rules for generating valid, high-quality skills

2. **Read and Use Template (REQUIRED)**
   - For English: Read [assets/templates/SKILL-TEMPLATE-EN.md](assets/templates/SKILL-TEMPLATE-EN.md)
   - For Chinese: Read [assets/templates/SKILL-TEMPLATE-ZH.md](assets/templates/SKILL-TEMPLATE-ZH.md)
   - Use the template structure as the foundation for the generated SKILL.md
   - Fill in all template sections with extracted workflow information

3. **Extract Information**
   - **Goal**: What the workflow accomplishes
   - **Prerequisites**: Required tools, permissions, or setup
   - **Steps**: Ordered sequence of actions
   - **Examples**: Concrete input/output from the conversation
   - **Notes**: Edge cases, warnings, or tips discovered

4. **Handle Single vs Multi-Feature Skills**
   - **Single Feature**: All steps and details go directly in SKILL.md
   - **Multi-Feature**: When the skill contains multiple distinct features/functions:
     - SKILL.md contains overview and links to each feature
     - Create separate documentation for each feature in `references/` directory
     - Example: `references/feature-one.md`, `references/feature-two.md`
     - SKILL.md should list all features with links: `See [Feature One](references/feature-one.md) for details`

5. **Abstract Specifics**
   - Replace hardcoded paths with `{{path}}` placeholders
   - Replace specific filenames with `{{filename}}` placeholders
   - Keep tool names and commands concrete

6. **Copy Tools (Use Bash cp command)**
   - If scripts or custom tools were used in the workflow, **MUST use COPY tool** to copy them to the new skill's `scripts/` directory
   - Example: `cp /path/to/original-script.sh {{skill-name}}/scripts/script.sh`
   - Update references in SKILL.md to use relative paths: `scripts/script-name.py`
   - Add an "Included Scripts" section in SKILL.md listing the bundled scripts/tools

7. **Validate Against Specification**
   - Ensure `name` field: 1-64 chars, lowercase, hyphens only
   - Ensure `description` field: 1-1024 chars, descriptive
   - Ensure frontmatter follows SPECIFICATION.md format
   - Ensure content follows BEST-PRACTICES.md guidelines
   - Keep main SKILL.md under 500 lines