skill_factory

🏭 Skill Factory

Core Mission: Standardized production, enabling ideas to be implemented quickly. This is the "mother" skill of the Yunshu system, responsible for incubating new capability modules.

1. Core Capabilities

1.1 Smart Scaffolding

• Automatic directory structure: Under skills/ , automatically create a folder structure compliant with TRAE specifications.
• Standardized metadata: Automatically generate name, description, tags, version containing SKILL.md。
• Dependency injection: Based on the skill type (such as Web Search or Python Script), automatically add package.json or requirements.txt template.

1.2 Templates

Template ID	Applicable Scenarios	Included Files
Basic	General simple tasks	SKILL.md
TypeScript	Web SDK integration (Search, VLM)	SKILL.md, scripts/*.ts, package.json
Python	Local data processing (PDF, Excel)	SKILL.md, scripts/*.py, requirements.txt
Agentic	Complex multi-step tasks (Novel Expert)	SKILL.md, guides/*.md, prompts/*.json

2. Context Parameters

This skill supports the following dynamic parameters, which can be automatically populated through the conversation context:

• {{skill_id}}: The unique identifier of the skill (in English, such as code_reviewer）。
• {{skill_name}}: The display name of the skill (such as 代码审查专家）。
• {{skill_type}}: Template type (basic | typescript | python | agentic）。
• {{description}}: A short description of the skill, used to generate Frontmatter.
• {{tags}}: The skill's tag list, used for context triggering.

3. Workflow

1. Requirement analysis: The user describes the skill they want to create (for example: "I want to make a skill that can automatically analyze Excel").
2. Parameter inference：
   • skill_id: excel_analyst
   • skill_type: python
   • tags: ['excel', 'data', 'analysis', 'python']
3. Confirmation and generation: After confirming the parameters with the user, call the file system tool to create the skill package.
4. Hot reload: Prompt the user to refresh the Skills list.

4. Best Practices

• Naming conventions：skill_id must be snake_case or kebab-case。
• Atomicity: A skill should do only one thing, keeping SKILL.md clear and focused.
• Documentation first: The generated SKILL.md must include detailed Overview, Usage, and Reference sections.