CM DocKit: Knowledge Systematization Engine

A professional knowledge systematization engine powered by codebase analysis and UX design principles. One source scan = one complete knowledge base — Personas, JTBD, Process Flows, Technical Docs, SOPs, API Reference. Supports plain Markdown output or a premium VitePress site. Includes SEO optimization, sitemap generation, and AI/LLM-readable content.

When to Activate

• User asks to "create documentation", "generate docs"
• User mentions "SOP", "user guide", "manual"
• User wants technical docs from a codebase
• User runs /DocKit Master

Document Types

Output Formats

Procedure

Step 1: Gather Input (Single Consolidated Prompt)

CRITICAL: Ask ALL questions in ONE message. Do NOT ask one at a time. Present the following intake form to the user, using this 📚 DocKit Master — Configuration

Please answer the following questions so I can automatically create an execution plan:

You can answer briefly, e.g.: "all, vitepress, full, yes, no, yes, yes"

🌏 Smart language rules:

1. Auto-detect: Determine default language from the user's chat language
   • User chats in Vietnamese → default vi
   • User chats in Chinese → default zh
   • User chats in Japanese → default ja
   • User chats in English → default en
   • (Applies similarly for any other language)
2. Multi-language (yes): Automatically add English (en) as secondary language
   • Example: Vietnamese user + multi-language → vi + en
   • Example: Chinese user + multi-language → zh + en
   • If user already chats in English + multi-language → ask which secondary language
3. Override: User can override by specifying explicitly, e.g.: "write in Japanese"

Step 1b: Auto-Generate Execution Plan

After receiving answers, immediately create an execution plan (do NOT ask more questions).

Map the answers to this execution config:

DOC_TYPE     = [knowledge | tech | sop | api | all]
FORMAT       = [markdown | vitepress]
SCOPE        = [full | focused]
FOCUS_TARGET = [directory/module name if focused, else null]
LANGUAGE     = [vi | en | vi+en]
I18N         = [yes | no] (only relevant for vitepress)
RECORD       = [yes | no]
PROJECT_PATH = [absolute path]
SEO          = [yes | no] (default: yes)
LLM_OPTIMIZE = [yes | no] (default: yes)

Then present the plan to user as a checklist artifact, like:

## 🚀 Execution Plan

- [ ] Scan code: [full/focused → target]
- [ ] Generate documents: [type] in [language]
- [ ] Export format: [markdown/vitepress]
- [ ] [If vitepress + i18n] Configure multi-language
- [ ] [If record] Record video walkthrough
- [ ] [If seo] Run SEO checklist + generate sitemap
- [ ] [If llm_optimize] Apply LLM optimization rules
- [ ] Review and deliver

After presenting the plan → proceed to Step 2 immediately (auto-execute). Do NOT wait for approval unless the plan has ambiguity.

Step 2: Analyze Codebase

Read and follow skills/analyze-codebase.md in this directory.

Output: structured analysis saved to docs/analysis.md (NOT _analysis.md) including:

• Project type, languages, frameworks
• Directory structure and architecture layers
• Entry points, routes, database schema
• Key business logic modules
• Dependencies overview
• Test coverage

Step 3: Apply Content Guidelines

MANDATORY — Read skills/content-guidelines.md before generating any content.

Key rules to enforce:

• Filenames: kebab-case, no underscores, no dots
• Frontmatter: Every .md file must have title, description, keywords, robots
• Quick Reference: Every doc starts with a summary box
• Progressive Disclosure: Use <details> for advanced content
• Admonitions: Use :::tip, :::info, :::warning, :::danger for callouts
• Mermaid: NO hardcoded colors — VitePress auto-adapts to light/dark
• Code Groups: Use :::code-group for multi-platform examples
• Internal Links: ≥2 cross-links per page

Step 3b: Apply SEO & LLM Guidelines (If enabled)

If SEO = yes: Read skills/content-writing.md for:

• Keyword placement (title, H1, first paragraph, H2s, meta)
• Inverted pyramid structure (answer first, details later)
• Active voice (≥80%), transition words (≥30%)
• FAQ in schema-ready format for rich snippets

If LLM_OPTIMIZE = yes: Read skills/llm-optimization.md for:

• Clean heading hierarchy (no skipped levels)
• Text descriptions alongside all Mermaid diagrams
• Self-contained sections (≤500 words per H2)
• Consistent terminology (glossary section in index)
• UTF-8 clean output

Step 4: Generate Documents

Based on the chosen type, read and follow the corresponding skill file:

• knowledge → Run 3 skills sequentially:

  1. Read skills/persona-builder.md → docs/personas/ (Buyer & User Personas)
  2. Read skills/jtbd-analyzer.md → docs/jtbd/ (JTBD Canvases)
  3. Read skills/flow-mapper.md → docs/flows/ (Workflow, Sequence, Lifecycle, Journey)

• tech → Read skills/tech-docs.md, generate:

  • docs/architecture.md — System architecture + ADR
  • docs/database.md — Database schema & data model
  • docs/deployment.md — Deployment & infrastructure
  • docs/data-flow.md — Data flow diagrams

• sop → Auto-run knowledge first if not yet generated, then:

  • Read skills/sop-guide.md, generate:
  • docs/sop/ — One .md per feature/module
  • Each file: Persona Context → Process Flow → Steps → Journey → Troubleshooting → FAQ

• api → Read skills/api-reference.md, generate:

  • docs/api/ — Organized by resource
  • Each file: Quick Ref → Endpoints table → Multi-language examples

• all → Run knowledge → tech → sop → api sequentially

Step 5: Export

Based on