coze-skill-creator

Coze Skill Creator

Task Goal

• This Skill is used to: create a complete Skill from a configuration file or requirement description, including tool configuration, workflow orchestration, and code generation
• Capabilities include: configuration validation, file generation, template rendering, and skill packaging
• Trigger conditions: the user describes skill requirements, provides a skill configuration JSON, or designs a workflow structure

Prerequisites

• Dependency description: the dependency packages and versions required by the scripts

  jsonschema>=4.0.0
  

• Non-standard file/folder preparation: none

Operation Steps

• Standard process:

  1. Requirement Analysis
     • Based on the user's description, the agent understands the skill's core capabilities, trigger scenarios, and tool requirements
     • Refer to references/config-reference.md to determine the configuration structure
  2. Configuration Generation
     • Based on the analysis results, generate the skill configuration JSON file
     • The configuration includes: metadata, tool list, workflow nodes, and script definitions
  3. Schema Validation
     • Call scripts/validate_schema.py --config <config_path> to validate the configuration format
     • If validation fails, correct the configuration based on the error message
  4. File Generation
     • Call scripts/generate_skill.py --config <config_path> --output <output_dir> to generate the skill files
     • The script automatically creates the SKILL.md, scripts/, references/, and assets/ structure
  5. Packaging and Testing
     • Use the package_skill(skill_dir_name=<skill_name>) tool to package the skill
     • Verify that the .skill file is generated successfully and its content is complete

• Optional branches:

  • When the user provides a configuration JSON: skip directly to step 3 for validation
  • When only a workflow design needs to be generated: refer to references/workflow-guide.md, and the agent generates the workflow diagram and node definitions
  • When only tool integration needs to be configured: refer to references/tool-reference.md, and the agent generates the tool configuration

Resource Index

• Required scripts:
  • scripts/validate_schema.py (purpose and parameters: validate whether the skill configuration JSON conforms to the Schema, parameters: --config)
  • scripts/generate_skill.py (purpose and parameters: generate complete skill files based on the configuration, parameters: --config, --output)
• Domain references:
  • references/config-reference.md (when to read: when generating or validating the configuration)
  • references/workflow-guide.md (when to read: when designing the workflow)
  • references/tool-reference.md (when to read: when configuring tools)
• Output assets:
  • assets/templates/skill-template.md (used directly to generate SKILL.md)
  • assets/templates/python-script.py (used directly to generate Python scripts)
  • assets/templates/schema.json (used to validate the configuration format)

Notes

• The configuration JSON must conform to the Schema defined by assets/templates/schema.json Defined Schema
• The generated skill name must use lowercase letters and hyphens; the -skill suffix is prohibited
• The connections between workflow nodes must maintain a directed acyclic graph structure
• Make full use of the agent's natural language understanding capabilities, and call scripts only when structured output is needed

Usage Examples

Example 1: Create a skill from a requirement description

• Function Description: The user describes "needs a skill to handle image compression"
• Execution method: The agent analyzes requirements → generates configuration → script verification → generates files → packaging
• Key parameters: skill name image-compressor, description, tool list
• Configuration example:

  {
    "name": "image-compressor",
    "description": "压缩图片文件大小",
    "tools": [
      {
        "name": "compress",
        "type": "python",
        "parameters": [{"name": "quality", "type": "integer"}]
      }
    ]
  }
  

Example 2: Generate a skill from a configuration file

• Function description: The user has provided the skill-config.json configuration file
• Execution method: script verification → script generation → packaging
• Key parameters:--config skill-config.json, --output ./my-skill
• Command:python scripts/generate_skill.py --config skill-config.json --output ./my-skill

Example 3: Design a workflow

• Function description: The user needs to create a skill workflow containing branching logic
• Execution method: The agent reads the workflow guide → designs nodes → generates configuration → generates the skill
• Key parameters: node types, connection relationships, conditional expressions
• Workflow example:

  [开始] → [条件判断] → [分支A: API调用] → [结束]
                    ↘ [分支B: 本地处理] → [结束]