generate-workflow-diagram
About
This skill generates themed Mermaid flowchart diagrams from `putior` workflow data. It offers multiple themes, output formats, and interactive features for embedding in documentation. Use it to visualize workflows after annotating source files or when you need to update diagrams for different audiences.
Quick Install
Claude Code
Recommendednpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/generate-workflow-diagramCopy and paste this command in Claude Code to install this skill
Documentation
Generate Workflow Diagram
Make themed Mermaid flowchart diagram from putior workflow data. Embed in docs.
When Use
- After annotating source files, ready to make visual diagram
- Regenerate diagram after workflow changes
- Switch themes or output formats for different audiences
- Embed workflow diagrams in README, Quarto, R Markdown docs
Inputs
- Required: Workflow data from
put(),put_auto(), orput_merge() - Optional: Theme name (default:
"light"; options: light, dark, auto, minimal, github, viridis, magma, plasma, cividis) - Optional: Output target: console, file path, clipboard, raw string
- Optional: Interactive features:
show_source_info,enable_clicks
Steps
Step 1: Extract Workflow Data
Get workflow data from one of three sources.
library(putior)
# From manual annotations
workflow <- put("./src/")
# From manual annotations, excluding specific files
workflow <- put("./src/", exclude = c("build-workflow\\.R$", "test_"))
# From auto-detection only
workflow <- put_auto("./src/")
# From merged (manual + auto)
workflow <- put_merge("./src/", merge_strategy = "supplement")
Workflow data frame may include node_type column from annotations. Node types control Mermaid shapes:
node_type | Mermaid Shape | Use Case |
|---|---|---|
"input" | Stadium ([...]) | Data sources, configuration files |
"output" | Subroutine [[...]] | Generated artifacts, reports |
"process" | Rectangle [...] | Processing steps (default) |
"decision" | Diamond {...} | Conditional logic, branching |
"start" / "end" | Stadium ([...]) | Entry/terminal nodes |
Each node_type also gets CSS class (class nodeId input;) for theme-based styling.
Got: Data frame with at least one row, has id, label, optionally input, output, source_file, node_type columns.
If fail: Data frame empty? No annotations or patterns found. Run analyze-codebase-workflow first, or check annotations syntactically valid with put("./src/", validate = TRUE).
Step 2: Pick Theme + Options
Pick theme for target audience.
# List all available themes
get_diagram_themes()
# Standard themes
# "light" — Default, bright colors
# "dark" — For dark mode environments
# "auto" — GitHub-adaptive with solid colors
# "minimal" — Grayscale, print-friendly
# "github" — Optimized for GitHub README files
# Colorblind-safe themes (viridis family)
# "viridis" — Purple→Blue→Green→Yellow, general accessibility
# "magma" — Purple→Red→Yellow, high contrast for print
# "plasma" — Purple→Pink→Orange→Yellow, presentations
# "cividis" — Blue→Gray→Yellow, maximum accessibility (no red-green)
Additional parameters:
direction: Diagram flow direction —"TD"(top-down, default),"LR"(left-right),"RL","BT"show_artifacts:TRUE/FALSE— show artifact nodes (files, data); noisy for large workflows (16+ extra nodes)show_workflow_boundaries:TRUE/FALSE— wrap each source file nodes in Mermaid subgraphsource_info_style: How source file info displayed on nodes (subtitle)node_labels: Format for node label text
Got: Theme names printed. Pick one by context.
If fail: Theme name not recognized? put_diagram() falls back to "light". Check spelling.
Step 3: Custom Palette with put_theme() (Optional)
9 built-in themes don't match project palette? Make custom theme with put_theme().
# Create custom palette — unspecified types inherit from base theme
cyberpunk <- put_theme(
base = "dark",
input = c(fill = "#1a1a2e", stroke = "#00ff88", color = "#00ff88"),
process = c(fill = "#16213e", stroke = "#44ddff", color = "#44ddff"),
output = c(fill = "#0f3460", stroke = "#ff3366", color = "#ff3366"),
decision = c(fill = "#1a1a2e", stroke = "#ffaa33", color = "#ffaa33")
)
# Use the palette parameter (overrides theme when provided)
mermaid_content <- put_diagram(workflow, palette = cyberpunk, output = "raw")
writeLines(mermaid_content, "workflow.mmd")
put_theme() takes input, process, output, decision, artifact, start, end node types. Each takes named vector c(fill = "#hex", stroke = "#hex", color = "#hex"). Unset types inherit from base theme.
Got: Mermaid output with custom classDef lines. Node shapes from node_type preserved; only colors change. All node types use stroke-width:2px — override not supported via put_theme().
If fail: Palette object not putior_theme class? put_diagram() raises descriptive error. Pass return value of put_theme(), not raw list.
Fallback — manual classDef replacement: Fine-grained control beyond put_theme() (per-type stroke widths)? Generate with base theme + replace classDef lines manually:
mermaid_content <- put_diagram(workflow, theme = "dark", output = "raw")
lines <- strsplit(mermaid_content, "\n")[[1]]
lines <- lines[!grepl("^\\s*classDef ", lines)]
custom_defs <- c(" classDef input fill:#1a1a2e,stroke:#00ff88,stroke-width:3px,color:#00ff88")
mermaid_content <- paste(c(lines, custom_defs), collapse = "\n")
Step 4: Generate Mermaid Output
Make diagram in desired output mode.
# Print to console (default)
cat(put_diagram(workflow, theme = "github"))
# Save to file
writeLines(put_diagram(workflow, theme = "github"), "docs/workflow.md")
# Get raw string for embedding
mermaid_code <- put_diagram(workflow, output = "raw", theme = "github")
# With source file info (shows which file each node comes from)
cat(put_diagram(workflow, theme = "github", show_source_info = TRUE))
# With clickable nodes (for VS Code, RStudio, or file:// protocol)
cat(put_diagram(workflow,
theme = "github",
enable_clicks = TRUE,
click_protocol = "vscode" # or "rstudio", "file"
))
# Full-featured
cat(put_diagram(workflow,
theme = "viridis",
show_source_info = TRUE,
enable_clicks = TRUE,
click_protocol = "vscode"
))
Got: Valid Mermaid code starts with flowchart TD (or LR by direction). Nodes connected by arrows showing data flow.
If fail: Output is flowchart TD with no nodes? Workflow data frame empty. Connections missing? Check output filenames match input filenames across nodes.
Step 5: Embed in Target Document
Insert diagram into appropriate docs format.
GitHub README (```mermaid code fence):
## Workflow
```mermaid
flowchart TD
A["Extract Data"] --> B["Transform"]
B --> C["Load"]
```
Quarto document (native mermaid chunk via knit_child):
# Chunk 1: Generate code (visible, foldable)
workflow <- put("./src/")
mermaid_code <- put_diagram(workflow, output = "raw", theme = "github")
# Chunk 2: Output as native mermaid chunk (hidden)
#| output: asis
#| echo: false
mermaid_chunk <- paste0("```{mermaid}\n", mermaid_code, "\n```")
cat(knitr::knit_child(text = mermaid_chunk, quiet = TRUE))
R Markdown (with mermaid.js CDN or DiagrammeR):
DiagrammeR::mermaid(put_diagram(workflow, output = "raw"))
Got: Diagram renders correct in target format. GitHub renders mermaid code fences native.
If fail: GitHub won't render diagram? Code fence must use exactly ```mermaid (no extra attributes). Quarto → use knit_child() approach since direct variable interpolation in {mermaid} chunks not supported.
Checks
-
put_diagram()produces valid Mermaid code (starts withflowchart) - All expected nodes appear in diagram
- Data flow connections (arrows) present between connected nodes
- Selected theme applied (check init block in output for theme-specific colors)
- Diagram renders correct in target format (GitHub, Quarto)
Pitfalls
- Empty diagrams: Usually
put()returned no rows. Check annotations exist + syntactically valid. - All nodes disconnected: Output filenames must exactly match input filenames (including extension) for putior to draw connections.
data.csv+Data.csvare different. - Theme not visible on GitHub: GitHub mermaid renderer has limited theme support.
"github"theme designed for GitHub.%%{init:...}%%theme block may be ignored by some renderers. - Quarto mermaid variable interpolation: Quarto
{mermaid}chunks don't support R variables direct. Useknit_child()from Step 5. - Clickable nodes not working: Click directives need renderer supporting Mermaid interaction events. GitHub static renderer no click support. Use local Mermaid renderer or putior Shiny sandbox.
- Self-referential meta-pipeline files: Scanning directory including build script generating diagram → duplicate subgraph IDs + Mermaid errors. Use
excludeparameter:workflow <- put("./src/", exclude = c("build-workflow\\.R$", "build-workflow\\.js$")) show_artifacts = TRUEtoo noisy: Large projects generate many artifact nodes (10–20+), clutter diagram. Useshow_artifacts = FALSE+ rely onnode_typeannotations to mark key inputs/outputs explicit.
See Also
annotate-source-files— prerequisite: files annotated before diagram generationanalyze-codebase-workflow— auto-detection supplements manual annotationssetup-putior-ci— automate diagram regeneration in CI/CDcreate-quarto-report— embed diagrams in Quarto reportsbuild-pkgdown-site— embed diagrams in pkgdown docs sites
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
