sf-diagram-mermaid

sf-diagram-mermaid: Salesforce Diagram Generation

Use this skill when the user needs text-based diagrams: Mermaid diagrams for architecture, OAuth, integration flows, ERDs, or Agentforce structure, plus ASCII fallback when plain-text compatibility matters.

When This Skill Owns the Task

Use sf-diagram-mermaid when the user wants:

• Mermaid output
• ASCII fallback diagrams
• architecture, sequence, flowchart, or ERD views in markdown-friendly form
• diagrams that can live directly in docs, READMEs, or issues

Delegate elsewhere when the user wants:

• rendered PNG/SVG images or polished mockups → sf-diagram-nanobananapro
• non-Salesforce systems only → use a more general diagramming skill
• object discovery before an ERD → sf-metadata

---

Supported Diagram Families

Type	Preferred Mermaid form	Typical use
OAuth / auth flows	sequenceDiagram	Authorization Code, JWT, PKCE, Device Flow
ERD / data model	flowchart LR	object relationships and sharing context
integration sequence	sequenceDiagram	request/response or event choreography
system landscape	flowchart	high-level architecture
role / access hierarchy	flowchart	users, profiles, permissions
Agentforce behavior map	flowchart	agent → topic → action relationships

---

Required Context to Gather First

Ask for or infer:

• diagram type
• scope and entities / systems involved
• output preference: Mermaid only, ASCII only, or both
• whether styling should be minimal, documentation-first, or presentation-friendly
• for ERDs: whether org metadata is available for grounding

---

Recommended Workflow

1. Pick the right diagram structure

• use sequenceDiagram for time-ordered interactions
• use flowchart LR for ERDs and capability maps
• keep a single primary story per diagram when possible

2. Gather data

For ERDs and grounded diagrams:

• use sf-metadata when real schema discovery is needed
• optionally use the local metadata helper script for counts / relationship context when appropriate

3. Generate Mermaid first

Apply:

• accurate labels
• simple readable node text
• consistent relationship notation
• restrained styling that renders cleanly in markdown viewers

4. Add ASCII fallback when useful

Provide an ASCII version when the user wants terminal compatibility or plaintext documentation.

5. Explain the diagram briefly

Call out the key relationships, flow direction, and any assumptions.

---

High-Signal Rules

For sequence diagrams

• use autonumber when step order matters
• distinguish requests vs responses clearly
• use notes sparingly for protocol detail

For ERDs

• prefer flowchart LR
• keep object cards simple
• use clear relationship arrows
• avoid field overload unless the user explicitly asks for field-level detail
• color-code object types only when it improves readability

For ASCII output

• keep width reasonable
• align arrows and boxes consistently
• optimize for readability over decoration

---

Output Format

## <Diagram Title>

### Mermaid Diagram
```mermaid
<diagram>
```

### ASCII Fallback
```text
<ascii>
```

### Notes
- <key point>
- <assumption or limitation>

---

Cross-Skill Integration

Need	Delegate to	Reason
real object / field definitions	sf-metadata	grounded ERD generation
rendered diagram / image output	sf-diagram-nanobananapro	visual polish beyond Mermaid
connected-app auth setup context	sf-connected-apps	accurate OAuth flows
Agentforce logic visualization	sf-ai-agentscript	source-of-truth behavior details
Flow behavior diagrams	sf-flow	actual Flow logic grounding

---

Reference Map

Start here

• references/diagram-conventions.md
• references/mermaid-reference.md
• references/usage-examples.md

Styling / ERD specifics

• references/mermaid-styling.md
• references/color-palette.md
• references/erd-conventions.md

Preview

• references/preview-guide.md
• scripts/mermaid_preview.py
• scripts/query-org-metadata.py

---

Score Guide

Score	Meaning
72–80	production-ready diagram
60–71	clear and useful with minor polish left
48–59	functional but could be clearer
35–47	needs structural improvement
< 35	inaccurate or incomplete