bp-visualize

name	bp-visualize
description	This skill should be used when the user asks to "visualize workflow", "generate diagram", "show workflow flow", "mermaid diagram", "graph workflow", "render workflow", "see workflow structure", or needs to create visual representations of workflow.yaml files. Triggers on "visualize", "diagram", "mermaid", "show flow", "graph", "flowchart", "render workflow", "workflow diagram".
allowed-tools	Read, Glob, AskUserQuestion

Visualize Workflow

Generate Mermaid diagrams from workflow.yaml files, supporting end-to-end flowcharts, subflow diagrams, and state diagrams.

Diagram Element Mapping: patterns/diagram-element-mapping.md Mermaid Generation Pattern: ${CLAUDE_PLUGIN_ROOT}/lib/patterns/mermaid-generation.md

---

Overview

This skill generates Mermaid diagrams from workflow.yaml files:

• End-to-end flowcharts -- Complete workflow visualization
• Subflow diagrams -- Selected portions of the workflow
• State diagrams -- Alternative representation for state machines

---

Prerequisites

Check these dependencies before execution:

Tool	Required	Check	Purpose
jq	Mandatory	command -v jq	JSON processing
yq	Mandatory	command -v yq	YAML processing
gh	Recommended	command -v gh	GitHub API access (private repos)

If mandatory tools are missing, exit with error and installation guidance.

---

Visualization Mappings

Node Types to Mermaid Shapes

Node Type	Mermaid Shape	Syntax
action	Rectangle	node[Action Name]
conditional	Diamond	node{Condition?}
conditional (audit)	Diamond	node{Validations?}
user_prompt	Stadium	node([User Prompt])
ending (success)	Rounded	node(Success)
ending (error)	Rounded	node(Error)

Transitions to Edge Styles

Transition	Style	Example
on_success	Solid green	`A -->
on_failure	Dashed red	`A -.->
branches.true	Solid	`A -->
branches.false	Dashed	`A -.->
on_response	Labeled	`A -->

---

Phase 1: Mode Detection

Parse any flags provided by the user or calling skill to determine visualization mode upfront.

Step 1.1: Parse Flags

Check for flags in the user's request or passed context:

Flag	Effect	Example
--format <type>	Set diagram format	--format flowchart, --format state, --format subflow
--output <path>	Set file output path	--output docs/workflow-diagram.md

Store parsed values:

computed:
  mode:
    format: "flowchart"   # flowchart | state | subflow (null if not provided)
    output_path: null      # file path or null for display-only

Step 1.2: Apply Mode or Prompt

• If --format provided: set computed.mode.format and skip the visualization mode prompt in Phase 3 (Step 3.3).
• If --output provided: set computed.mode.output_path and skip the file output prompt in Phase 5 (Step 5.2).
• If no flags provided: proceed normally -- the user will be prompted for format choice in Phase 3.

---

Phase 2: Load Workflow

Step 2.1: Determine Target Workflow

If user provided a path:

1. Validate the path exists
2. Read the workflow.yaml file
3. Store in computed.workflow_content

If no path provided:

1. Check if current directory has a workflow.yaml:
   • Glob for **/workflow.yaml in current directory
   • If found, ask user which one to visualize
2. If none found, ask user for the path:

   {
     "questions": [{
       "question": "Which workflow would you like to visualize?",
       "header": "Target",
       "multiSelect": false,
       "options": [
         {"label": "Provide path", "description": "I'll give you the workflow.yaml path"},
         {"label": "Search current directory", "description": "Look for workflow.yaml files here"},
         {"label": "Use current skill", "description": "Visualize the skill I'm currently in"
       
     
   
   

Step 2.2: Parse Workflow Structure

Extract from workflow.yaml:

• name -- Workflow identifier
• start_node -- Entry point
• nodes -- All workflow nodes with types and transitions
• endings -- Terminal states

Store in:

computed:
  workflow_name: "extracted name"
  start_node: "check_arguments"
  nodes:
    - id: "node_id"
      type: "action|conditional|user_prompt|..."
      description: "Node description"
      transitions: {...}
  endings:
    - id: "success"
      type: "success|error"

---

Phase 3: Analyze & Choose Mode

Step 3.1: Detect Decision Nodes

Identify nodes that create branching:

• conditional nodes -- branches.true and branches.false
• conditional (audit) nodes -- Same as conditional but validates all conditions
• user_prompt nodes -- Multiple on_response handlers

Count and classify:

computed:
  decision_nodes:
    - id: "check_arguments"
      type: "conditional"
      branch_count: 2
    - id: "show_main_menu"
      type: "user_prompt"
      branch_count: 4

Step 3.2: Identify Subflow Boundaries

Trace workflow paths to detect natural segments:

1. Trace from start: Follow paths from start_node to endings
2. Identify clusters: Group nodes that form logical units
3. Name subflows: Use node descriptions or infer from purpose

Algorithm:

For each decision node:
  1. Trace forward to next decision or ending
  2. Collect all nodes in this segment
  3. Name based on entry node description
  4. Count nodes in segment

Example detected subflows:

computed:
  subflows:
    - name: "Intent Parsing Flow"
      entry: "check_arguments"
      nodes: ["check_arguments", "parse_intent", "check_clear_winner"]
      exit_to: ["execute_matched_intent", "show_disambiguation"]
    - name: "Menu Flow"
      entry: "show_main_menu"
      nodes: ["show_main_menu", "delegate_init", "delegate_discover"]
      exit_to: ["success", "error_delegation"]

Step 3.3: Present Visualization Options

Ask user which visualization mode:

{
  "questions": [{
    "question": "How would you like to visualize the workflow?",
    "header": "Mode",
    "multiSelect": false,
    "options": [
      {"label": "Full diagram (Recommended)", "description": "Complete end-to-end workflow visualization"},
      {"label": "Subflows", "description": "Choose specific portions to visualize"},
      {"label": "State diagram", "description": "Alternative stateDiagram representation"}
    ]
  }]
}

If Subflows selected, present detected subflows:

{
  "questions": [{
    "question": "Which subflows would you like to visualize?",
    "header": "Subflows",
    "multiSelect": true,
    "options": [
      {"label": "Intent Parsing", "description": "check_arguments -> check_clear_winner (3 nodes)"},
      {"label": "Menu Flow", "description": "show_main_menu -> delegations (8 nodes)"},
      {"label": "Disambiguation", "description": "show_disambiguation -> delegations (5 nodes)"},
      {"label": "Error Handling",  
    
  

Store computed.diagram_mode ("flowchart" or "state"), computed.selected_nodes, and computed.orientation ("TD" or "LR") based on selection. Use TD for deep workflows (4+ decision nodes), LR for shallow workflows or subflow views.

---

Phase 4: Generate Diagram

Step 4.1: Build Node Definitions

For each node in the selected scope, generate Mermaid syntax:

For each node:
  shape = get_shape(node.type)
  label = sanitize(node.description or node.id)
  output: "{node.id}{shape_open}{label}{shape_close}"

Shape mapping:

action         -> [Label]
conditional    -> {Label?}
user_prompt    -> ([Label])
ending.success -> (Success)
ending.error   -> (Error)

Step 4.2: Build Edges

For each node, generate transition edges:

Action nodes:

{node_id} -->|success| {on_success}
{node_id} -.->|failure| {on_failure}

Conditional nodes:

{node_id} -->|true| {branches.true}
{node_id} -.->|false| {branches.false}

User prompt nodes:

{node_id} -->|{option_id}| {on_response.{option_id}.next_node}

Only emit edges whose targets are within computed.selected_nodes or ending IDs.

Step 4.3: Apply Styling

Add class definitions for visual distinction:

classDef success fill:#90EE90,stroke:#228B22,color:#000
classDef error fill:#FFB6C1,stroke:#DC143C,color:#000
classDef conditional fill:#87CEEB,stroke:#4682B4,color:#000
classDef userPrompt fill:#DDA0DD,stroke:#9932CC,color:#000

Apply classes to nodes:

• Endings with type: success -- :::success
• Endings with type: error -- :::error
• Conditional nodes -- :::conditional
• User prompt nodes -- :::userPrompt

Step 4.4: Assemble Complete Diagram

Combine into final output:

For flowchart mode: header (flowchart TD|LR), classDefs, start connector (start([Start]) --> {start_node}), node definitions, edges.

For state mode: header (stateDiagram-v2), start transition ([*] --> {start_node}), state transitions with colon-labeled edges, terminal transitions ({success} --> [*]).

Store assembled output in computed.diagram_output.

---

Phase 5: Output

Step 5.1: Display Diagram

Output the Mermaid diagram in a code block:

```mermaid
flowchart TD
    classDef success fill:#90EE90,stroke:#228B22,color:#000
    classDef error fill:#FFB6C1,stroke:#DC143C,color:#000
    classDef conditional fill:#87CEEB,stroke:#4682B4,color:#000
    classDef userPrompt fill:#DDA0DD,stroke:#9932CC,color:#000

    start([Start]) --> check_arguments{has_arguments?}
    check_arguments -->|true| parse_intent[Parse Intent]
    check_arguments -->|false| show_menu([Show Menu])
    ...
```

Step 5.2: Offer File Output

Ask if user wants to save:

{
  "questions": [{
    "question": "Would you like to save this diagram to a file?",
    "header": "Save",
    "multiSelect": false,
    "options": [
      {"label": "Display only", "description": "Just show it here"},
      {"label": "Save to docs/", "description": "Write to docs/workflow-diagram.md"},
      {"label": "Save to README", "description": "Append to README.md"}
    ]
  }]
}

If saving:

1. Wrap diagram in appropriate markdown
2. Write to specified location
3. Confirm save path

---

Output Examples

End-to-End Flowchart

flowchart TD
    classDef success fill:#90EE90,stroke:#228B22,color:#000
    classDef error fill:#FFB6C1,stroke:#DC143C,color:#000
    classDef conditional fill:#87CEEB,stroke:#4682B4,color:#000
    classDef userPrompt fill:#DDA0DD,stroke:#9932CC,color:#000

    start([Start]) --> check_arguments{has_arguments?}:::conditional
    check_arguments -->|true| parse_intent[Parse Intent]
    check_arguments -->|false| show_menu([Show Menu]):::userPrompt
    parse_intent --> check_winner{clear_winner?}:::conditional
    check_winner -->|true| execute[Execute Intent]
    check_winner -->|false| disambiguate([Disambiguate]):::userPrompt
    execute --> success_end(Success):::success
    show_menu --> delegate[Delegate to Skill]
    delegate --> success_end
    delegate -.->|failure| error_end(Error):::error

Subflow (Selected Nodes Only)

flowchart LR
    parse_intent[Parse Intent] --> check_winner{clear_winner?}:::conditional
    check_winner -->|true| execute[Execute Intent]
    check_winner -->|false| disambiguate([Disambiguate]):::userPrompt

State Diagram (Alternative)

stateDiagram-v2
    [*] --> check_arguments
    check_arguments --> parse_intent: has arguments
    check_arguments --> show_main_menu: no arguments
    parse_intent --> check_clear_winner
    check_clear_winner --> execute_matched_intent: clear winner
    check_clear_winner --> show_disambiguation: ambiguous
    execute_matched_intent --> success
    show_main_menu --> delegate_init: init
    show_main_menu --> delegate_discover: discover
    delegate_init --> success
    delegate_discover --> success
    success --> [*]

---

Reference Documentation

• Diagram Element Mapping: patterns/diagram-element-mapping.md
• Mermaid Generation Pattern: ${CLAUDE_PLUGIN_ROOT}/lib/patterns/mermaid-generation.md
• Node Mapping Pattern: ${CLAUDE_PLUGIN_ROOT}/lib/patterns/node-mapping.md

---

Related Skills

• Skill assessment: ${CLAUDE_PLUGIN_ROOT}/skills/bp-assess/SKILL.md
• Workflow extraction: ${CLAUDE_PLUGIN_ROOT}/skills/bp-extract/SKILL.md
• Skill building: ${CLAUDE_PLUGIN_ROOT}/skills/bp-build/SKILL.md
• Skill maintenance: ${CLAUDE_PLUGIN_ROOT}/skills/bp-maintain/SKILL.md