twwch/workflow-skill/coze-workflow

Name: coze-workflow
Author: twwch

Claude Code and agent skill by twwch/workflow-skill

Trust tier 3

community

Source check pending

Checking that the SKILL.md file is still in the repository.

Generate Coze workflow YAML files packaged as ZIP for import into coze.cn cloud platform. Produces importable workflow definitions with correct node schemas, edges, and layout.

Install

npx vskill@latest install twwch/workflow-skill/coze-workflow

SKILL.md

Coze Workflow Generator

Overview

This skill generates Coze workflow files that can be directly imported into coze.cn cloud platform via ZIP upload. Given a natural language description of a desired workflow, it produces a complete YAML file packaged as a ZIP with correct node schemas, edges, layout, and byte-level zip formatting.

The skill triggers when the user asks to:

Create, generate, or build a Coze workflow
Convert a process description into a Coze workflow
Scaffold a workflow for a Coze bot or application

Primary output: .zip file for coze.cn cloud import (via UI "导入" dialog).

Smart Interaction Logic

Before generating, assess whether the user's description is sufficient:

Proceed directly if the description includes:

Clear input/output expectations
Processing logic (what the workflow should do)
Enough detail to select appropriate nodes

Ask clarifying questions (max 3 rounds) if unclear:

"What inputs will the workflow receive, and what outputs should it produce?"
"What processing steps are needed? (e.g., LLM call, loop iteration, data aggregation)"
"Any specific models or plugins to use?"

Node Type Reference

Core node types (all verified from real coze.cn export)

YAML Type	Purpose
`start`	Entry point; defines workflow input variables
`end`	Terminal node; collects outputs
`llm`	Invokes a large language model
`loop`	Iterates over array with inner sub-nodes
`plugin`	Executes a registered plugin/tool
`variable_merge`	Merges variables from multiple branches
`image_generate`	Generates images from prompts
`code`	Run Python (language: 3) / TypeScript (language: 5)
`http`	HTTP API calls (GET/POST/PUT/DELETE)
`condition`	Conditional branching (IF/ELIF/ELSE)
`intent`	LLM-based intent classification
`knowledge`	Knowledge base search
`text`	Text concat/split
`question`	Pause for user input
`batch`	Parallel array processing
`subflow`	Sub-workflow call

Additional node types (verified from real coze.cn export)

YAML Type	Purpose	Builder Function
`code`	Run Python/TypeScript code	`code_node()`
`http`	HTTP API calls	`http_node()`
`condition`	Conditional branching (IF/ELSE)	`selector_node()`
`intent`	LLM-based intent classification	`intent_node()`
`knowledge`	Knowledge base search	`knowledge_node()`
`subflow`	Sub-workflow call	—
`batch`	Parallel array processing	—
`text`	Text concat/split	—
`question`	Pause for user input	—
`to_json`	JSON serialize	—
`from_json`	JSON deserialize	—
`assign_variable`	Variable assignment	—
`database`	SQL custom query	—
`insert_record` / `update_record` / `select_record` / `delete_record`	DB CRUD	—
`dataset_write` / `dataset_delete`	Knowledge write/delete	—
`image_generate` / `drawing_board`	Image generation	—
`video_generation` / `video_audio_extractor` / `video_frame_extractor`	Video	—
`conversation_create` / `conversation_update` / `conversation_delete` / `conversation_list`	Conversation mgmt	—
`conversation_history_list` / `conversation_clear`	Conversation history	—
`message_create` / `message_update` / `message_delete` / `message_list`	Message mgmt	—
`ltm_write` / `ltm_read`	Long-term memory	—

Condition edge ports: sourcePortID="true" (IF), sourcePortID="false" (ELSE), sourcePortID="true_1" (ELIF) Intent edge ports: sourcePortID="branch_0", sourcePortID="branch_1", ..., sourcePortID="default"

YAML Format (coze.cn Cloud)

Critical formatting rules (verified via byte-level experiments)

4-space indentation for all nesting levels
Double quotes for string IDs: id: "100001", not id: '100001'
YAML must NOT be generated via `yaml.dump()` — use string template concatenation to preserve exact formatting
Node IDs: Start = "100001", End = "900001", others = 6-digit strings

Top-level structure

yaml

schema_version: 1.0.0
name: workflow_name
id: 7585079438426600001
description: "Description here"
mode: workflow
icon: plugin_icon/workflow.png
nodes:
    - id: "100001"
      type: start
      ...
    - id: "200001"
      type: llm
      ...
    - id: "900001"
      type: end
      ...
edges:
    - source_node: "100001"
      target_node: "200001"
    - source_node: "200001"
      target_node: "900001"

Data references (between nodes)

yaml

value:
    path: output_field_name
    ref_node: "source_node_id"

Start node template

yaml

    - id: "100001"
      type: start
      title: 开始
      icon: https://lf3-static.bytednsdoc.com/obj/eden-cn/dvsmryvd_avi_dvsm/ljhwZthlaukjlkulzlp/icon/icon-Start-v2.jpg
      description: "工作流的起始节点，用于设定启动工作流需要的信息"
      position:
        x: -1810
        y: 0
      parameters:
        node_outputs:
            input_var_name:
                type: string
                required: true
                value: null

LLM node template (complete — all 14 llmParam fields required)

yaml

    - id: "200001"
      type: llm
      title: Node Title
      icon: https://lf3-static.bytednsdoc.com/obj/eden-cn/dvsmryvd_avi_dvsm/ljhwZthlaukjlkulzlp/icon/icon-LLM-v2.jpg
      description: "调用大语言模型,使用变量和提示词生成回复"
      version: "3"
      position:
        x: 0
        y: 0
      parameters:
        fcParamVar:
            knowledgeFCParam: {}
        llmParam:
            - name: apiMode
              input:
                type: integer
                value: "0"
            - name: maxTokens
              input:
                type: integer
                value: "4096"
            - name: spCurrentTime
              input:
                type: boolean
                value: false
            - name: spAntiLeak
              input:
                type: boolean
                value: false
            - name: responseFormat
              input:
                type: integer
                value: "2"
            - name: modelName
              input:
                type: string
                value: Kimi-K2-250905
            - name: modelType
              input:
                type: integer
                value: "1763350148"
            - name: generationDiversity
              input:
                type: string
                value: balance
            - name: parameters
              input:
                type: object
                value: null
            - name: prompt
              input:
                type: string
                value: "{{input}}"
            - name: enableChatHistory
              input:
                type: boolean
                value: false
            - name: chatHistoryRound
              input:
                type: integer
                value: "3"
            - name: systemPrompt
              input:
                type: string
                value: "Your system prompt here"
            - name: stableSystemPrompt
              input:
                type: string
                value: ""
            - name: canContinue
              input:
                type: boolean
                value: false
            - name: loopPromptVersion
              input:
                type: string
                value: ""
            - name: loopPromptName
              input:
                type: string
                value: ""
            - name: loopPromptId
              input:
                type: string
                value: ""
        node_inputs:
            - name: input
              input:
                type: string
                value:
                    path: output
                    ref_node: "100001"
        node_outputs:
            output:
                type: string
                value: null
        settingOnError:
            processType: 1
            retryTimes: 0
            switch: false
            timeoutMs: 180000

End node template

yaml

    - id: "900001"
      type: end
      title: 结束
      icon: https://lf3-static.bytednsdoc.com/obj/eden-cn/dvsmryvd_avi_dvsm/ljhwZthlaukjlkulzlp/icon/icon-End-v2.jpg
      description: "工作流的最终节点，用于返回工作流运行后的结果信息"
      position:
        x: 1810
        y: 0
      parameters:
        node_inputs:
            - name: output
              input:
                value:
                    path: output
                    ref_node: "200001"
        terminatePlan: returnVariables

Loop node template (with inner sub-nodes)

yaml

    - id: "200002"
      type: loop
      title: Loop Title
      icon: https://lf3-static.bytednsdoc.com/obj/eden-cn/dvsmryvd_avi_dvsm/ljhwZthlaukjlkulzlp/icon/icon-Loop-v2.jpg
      description: "用于通过设定循环次数和逻辑，重复执行一系列任务"
      position:
        x: 0
        y: 0
      canvas_position:
        x: -200
        y: 200
      parameters:
        loopCount:
            type: integer
            value:
                content: 10
                rawMeta:
                    type: 2
                type: literal
        loopType: array
        node_inputs:
            - name: input
              input:
                type: list
                items:
                    type: string
                    value: null
                value:
                    path: output
                    ref_node: "200001"
        node_outputs:
            output:
                value:
                    type: list
                    items:
                        type: string
                        value: null
                    value:
                        path: output
                        ref_node: "last_inner_node_id"
        variableParameters: []
      nodes:
        - id: "301001"
          type: llm
          ...inner LLM node (8-space indent)...
      edges:
        - source_node: "200002"
          target_node: "301001"
          source_port: loop-function-inline-output
        - source_node: "301001"
          target_node: "200002"
          target_port: loop-function-inline-input

Loop edge rules:

Entry edge: source_node: "<loop_id>", source_port: loop-function-inline-output → first inner node
Exit edge: last inner node → target_node: "<loop_id>", target_port: loop-function-inline-input
Top-level edge out of loop: source_port: loop-output

Variable Merge node template

yaml

    - id: "200010"
      type: variable_merge
      title: 变量聚合
      icon: https://lf3-static.bytednsdoc.com/obj/eden-cn/dvsmryvd_avi_dvsm/ljhwZthlaukjlkulzlp/icon/VariableMerge-icon.jpg
      description: "对多个分支的输出进行聚合处理"
      position:
        x: 500
        y: 0
      parameters:
        mergeGroups:
            - name: v
              variables:
                - type: string
                  value:
                    content:
                        blockID: "200003"
                        name: output
                        source: block-output
                    rawMeta:
                        type: 1
                    type: ref
        node_outputs:
            output:
                type: string
                value: null

Edge format

yaml

edges:
    - source_node: "100001"
      target_node: "200001"
    - source_node: "200001"
      target_node: "200002"
      source_port: loop-output

ZIP Packaging

Directory structure (naming contract — strictly enforced)

Workflow-<NAME>-draft-<DIGITS>/
├── MANIFEST.yml
└── workflow/
    └── <NAME>-draft.yaml

<NAME> must match in directory name, filename, and YAML name: field
<NAME> must be ^[a-zA-Z][a-zA-Z0-9_]*$
<DIGITS> can be any number (e.g., 0001)

MANIFEST.yml format

This is the start of the file. Read the full SKILL.md on GitHub.

Security

Security scanPassed, 90 of 100Pattern checks on the skill's files.
Model reviewPassed

See the full security report

Quality evaluation

Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

See evaluation results

Source

SKILL.md in the repositoryCheckingskills/coze-workflow/SKILL.md

Popularity

Repository stars: 56
Repository forks: 14
7-day trend score: 20

Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 26 Apr 2026, 6 months ago.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

Claude Code
Cursor
GitHub Copilot
Windsurf
Codex
Gemini CLI
Cline
Amp
Roo Code
Goose
Continue
Kilo Code
Junie
Trae
OpenHands
Qwen Code
AI Agent
Tabnine Chat
Sourcegraph Cody
Amazon Q
Firebase Genkit
Sweep
Devin
Phind
Mentat
Claude Desktop
Aider
GPT Pilot
Plandex
bolt.new
v0
Replit Agent
Anse
TypingMind
BetterChatGPT
LibreChat
Warp AI
Pieces
Fig AI

Embed badge

[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/twwch/workflow-skill/coze-workflow/badge)](https://verified-skill.com/skills/twwch/workflow-skill/coze-workflow)

Something wrong with this skill or its page? Tell us.