Document Generator

Automatically Generate SOPs, READMEs, API Docs, and Runbooks

Documentation is always out of date, never complete, and takes hours to write well. The Document Generator takes your source material — code, notes, descriptions, or meeting transcripts — and generates properly structured, audience-appropriate documentation in minutes. SOPs, READMEs, API docs, runbooks, tech specs, and meeting notes.

What It Does

Trigger this workflow with a source type, source content, and target document type. Here's what happens:

1. Source Reading: The workflow reads your source material based on the type you specify:

• File: reads a single file (code, notes, etc.)
• Directory: scans a code directory, reads up to 20 files (TypeScript, JavaScript, Python, Go), extracts the first 100 lines of each
• URL: fetches the content of a web page
• Text: uses raw text you provide directly as input
• Transcript: reads a meeting or interview transcript

2. Content Analysis: Claude analyzes the source material to understand: what it does, how it works, who uses it, key inputs and outputs, edge cases and gotchas, prerequisites, and the most important concepts a reader needs to understand.

3. Document Generation: Based on the requested document type, Claude generates appropriately structured content:

• README: Project overview, installation, quick start, configuration, usage examples, contributing guide
• SOP: Step-by-step procedure with numbered steps, decision points, troubleshooting, escalation paths
• API Docs: Endpoint descriptions, request/response schemas, authentication, error codes, code examples
• Runbook: Incident response procedures, diagnostic steps, recovery actions, escalation contacts
• Tech Spec: Problem statement, proposed solution, alternatives considered, implementation plan, open questions
• Meeting Notes: Key decisions, action items with owners, follow-ups, next meeting agenda

4. Audience Adaptation: The technical depth, assumed knowledge, and language are calibrated based on your specified audience (technical, non-technical, or mixed).

5. Output Format: Delivered in your choice of markdown, Notion page format, Confluence format, or Google Docs format.

Who This Is For

Engineers who need to document code they've already written, team leads generating SOPs for repeated processes, and anyone who captures information in one format (code, transcripts, notes) and needs to turn it into polished documentation.

What's Included

• document-generator.yml — complete workflow with manual trigger, multi-source reading, AI content analysis, audience-aware generation, and multi-format output
• Support for 5 input types (file, directory, URL, text, transcript)
• 6 document types (README, SOP, API docs, runbook, tech spec, meeting notes)
• 3 audience levels (technical, non-technical, mixed)
• 4 output formats (markdown, Notion, Confluence, Google Docs)

Setup Steps

1. Install the workflow — copy document-generator.yml into your OpenClaw workspace
2. Run it — trigger with: source type, source path/content, document type, audience, and output format
3. Review the output — generated documents are ready to use with minimal editing

No ongoing configuration required. This is a pure on-demand workflow.

Example: Documenting a Code Directory

openclaw workflow run document-generator \
  --source-type directory \
  --source ./src/api \
  --doc-type api-docs \
  --audience technical \
  --output-format markdown

The workflow reads your API source files, infers endpoints, request/response shapes, and auth patterns, and generates complete API documentation.

Why This Beats Writing Docs Manually

Writing good documentation for an existing codebase takes 2–8 hours depending on complexity. The cognitive load comes from switching between reading code (technical thinking) and writing prose for humans (communication thinking). This workflow handles the cognitive switching — you get a complete, structured draft that needs editing rather than a blank page.

The quality improvement over time: as you use the workflow, you get a sense of what level of source detail produces the best output. A well-commented codebase generates excellent API docs. A detailed meeting transcript generates precise action items.

Result: documentation that actually gets written, on a timeline that doesn't compete with shipping.