Workflows Documentation

Documentation Creation Specialist - Unified Markdown & Component Management

Unified specialist providing: (1) Document quality pipeline with structure enforcement and content optimization, (2) OpenCode component creation (skills, agents, commands) with scaffolding, validation, and packaging, (3) ASCII flowchart creation for visualizing workflows, and (4) Install guide creation for setup documentation.

Core Principle: Structure first, then content, then quality.

Architecture: Scripts handle deterministic parsing/metrics, AI handles quality judgment and recommendations.

1. 🎯 WHEN TO USE

Use Case: Document Quality Management

Enforce markdown structure, optimize content for AI assistants, validate quality through script-assisted AI analysis.

README Creation - Use readme_template.md when:

• Creating new README for any folder or project
• User requests "create a README", "add documentation"
• Folder needs comprehensive documentation

Frontmatter Validation - Use frontmatter_templates.md when:

• Validating YAML frontmatter in any document
• Checking required fields for document types
• Fixing frontmatter syntax errors

Validation Workflow - Apply after Write/Edit operations:

• Auto-correct filename violations (ALL CAPS → lowercase, hyphens → underscores)
• Fix safe violations (separators, H2 case, emoji per rules)
• Check critical violations (missing frontmatter, wrong section order)

Manual Optimization - Run when:

• README needs optimization for AI assistants
• Creating critical documentation (specs, knowledge, skills)
• Pre-release quality checks
• Generating llms.txt for LLM navigation

Use Case: OpenCode Component Creation

Create and manage OpenCode components: skills, agents, and commands. Each component type has templates, validation, and quality standards.

Component Types:

• Skills (.opencode/skill/) - Knowledge bundles with workflows → skill_creation.md
• Agents (.opencode/agent/) - AI personas with tool permissions → agent_template.md
• Commands (.opencode/command/) - Slash commands for user invocation → command_template.md

Use when:

• User requests skill creation ("create a skill", "make a new skill")
• User requests agent creation ("create an agent", "make a new agent")
• User requests command creation ("create a command", "add a slash command")
• Scaffolding component structure
• Validating component quality
• Packaging skill for distribution

Skill Process (6 steps): Understanding (examples) → Planning (resources) → Initialization (init_skill.py) → Editing (populate) → Packaging (package_skill.py) → Iteration (test/improve)

Agent Process: Load agent_template.md → Define frontmatter (tools, permissions) → Create sections (workflow, capabilities, anti-patterns) → Validate → Test

Command Process: Load command_template.md → Define frontmatter (name, description) → Create execution logic → Add to command registry → Test

Use Case: Flowchart Creation

Create ASCII flowcharts for visualizing workflows, user journeys, and decision trees.

Use when:

• Documenting multi-step processes with branching
• Creating decision trees with multiple outcomes
• Showing parallel execution with sync points
• Visualizing approval gates and revision cycles

See: assets/flowcharts/

Use Case: Install Guide Creation

Create and validate installation documentation for MCP servers, plugins, and tools using phase-based templates.

Use when:

• Creating documentation for MCP server installation
• Documenting plugin setup procedures
• Standardizing tool installation across platforms
• Need phase-based validation checkpoints

5-Phase Process: Overview → Prerequisites → Installation → Configuration → Verification

See: install_guide_standards.md

When NOT to Use (All Modes)

• Non-markdown files (only .md supported)
• Simple typo fixes (use Edit tool directly)
• Internal notes or drafts
• Auto-generated API docs
• Very simple 2-3 step processes (use bullet points)
• Code architecture (use mermaid diagrams)

2. 🧭 SMART ROUTING & REFERENCES

Mode Selection

TASK CONTEXT
    │
    ├─► Improving markdown / documentation quality
    │   └─► MODE 1: Document Quality
    │       ├─► README creation: readme_template.md
    │       ├─► Knowledge files, general markdown
    │       ├─► Frontmatter validation: frontmatter_templates.md
    │       ├─► Quality analysis: extract_structure.py → JSON
    │       └─► llms.txt generation (ask first)
    │
    ├─► Creating OpenCode components (skills, agents, commands)
    │   └─► MODE 2: Component Creation
    │       ├─► Skills: init_skill.py + skill_md_template.md
    │       ├─► Agents: agent_template.md
    │       └─► Commands: command_template.md
    │
    ├─► Creating ASCII flowcharts / diagrams
    │   └─► MODE 3: Flowchart Creation
    │       └─► Load flowchart assets by pattern type
    │
    ├─► Creating install guide / setup documentation
    │   └─► MODE 4: Install Guide Creation
    │       └─► Load: install_guide_template.md
    │
    └─► Quick reference / standards lookup
        └─► Load: quick_reference.md

Resource Router

Mode 1 - Document Quality:

Mode 2 - OpenCode Component Creation:

Mode 3 - Flowcharts:

Mode 4 - Install Guide Creation:

General Utilities:

Core References

Templates

Resource Router

def route_documentation_resources(task):
    """Route to appropriate documentation resources."""
    
    # Mode 1: Skill Creation
    if task.involves("skill creation") or task.involves("new skill"):
        load("references/skill_creation.md")
        load("assets/opencode/skill_md_template.md")
        return "Mode 1: Skill Creation"
    
    if task.involves("reference file") or task.involves("bundled resource"):
        load("assets/opencode/skill_reference_template.md")
        return "Mode 1: Reference Creation"
    
    # Mode 2: Document Quality
    if task.involves("DQI") or task.involves("quality score"):
        load("references/validation.md")
        return "Mode 2: Document Quality"
    
    # Mode 3: Content Optimization
    if task.involves("optimize") or task.involves("AI context"):
        load("references/optimization.md")
        return "Mode 3: Content Optimization"
    
    # Mode 4: Flowchart Creation
    if task.involves("flowchart") or task.involves("ASCII diagram"):
        load("assets/flowcharts/")
        return "Mode 4: Flowchart Creation"
    
    # Mode 5: Install Guide
    if task.involves("install guide") or task.involves("setup instructions"):
        load("assets/documentation/install_guide_template.md")
        return "Mode 5: Install Guide"
    
    # Default
    load("references/quick_reference.md")
    return "Quick Reference"

Key Insight: Always run extract_structure.py first - it provides the structured JSON that enables accurate AI quality assessment. Without it, quality evaluation is subjective guesswork.

3. 🛠️ HOW IT WORKS

Mode 1: Document Quality

Script-Assisted AI Analysis:

# 1. Extract document structure to JSON
scripts/extract_structure.py path/to/document.md

# 2. AI receives JSON with:
#    - Frontmatter, structure, metrics
#    - Checklist results, DQI score
#    - Evaluation questions

# 3. AI reviews and provides recommendations

Document Type Detection (auto-applies enforcement):

Mode 2: OpenCode Component Creation

Skill Creation

Progressive Disclosure Design:

1. Metadata (name + description) - Always in context (~100 words)
2. SKILL.md body - When skill triggers (<5k words)
3. Bundled resources - As needed (unlimited)

After packaging: Run extract_structure.py on SKILL.md for final quality review.

Typical Workflow:

# 1. Initialize skill structure
scripts/init_skill.py my-skill --path .opencode/skill

# 2. Edit SKILL.md and bundled resources
# [User populates templates with content]

# 3. Quick validation check
scripts/quick_validate.py .opencode/skill/my-skill --json

# 4. Package with full validation
scripts/package_skill.py .opencode/skill/my-skill

# 5. Quality assurance (DQI scoring)
scripts/extract_structure.py .opencode/skill/my-skill/SKILL.md

Agent Creation

Template-First Workflow:

1. Load agent_template.md for structure reference
2. Create agent file in .opencode/agent/
3. Define YAML frontmatter (name, tools, permissions)
4. Create required sections (workflow, capabilities, anti-patterns)
5. Validate frontmatter syntax
6. Test with real examples

Key Difference from Skills: Agents have tool permissions (true/false per tool) and action permissions (allow/deny), not just allowed-tools array.

Command Creation

Template-First Workflow:

1. Load command_template.md for structure reference
2. Create command file in .opencode/command/
3. Define YAML frontmatter (name, description, triggers)
4. Create execution logic and examples
5. Add to command registry
6. Test invocation

Mode 3: Flowchart Creation

Building Blocks:

Process Box:        Decision Diamond:     Terminal:
┌─────────────┐         ╱──────╲           ╭─────────╮
│   Action    │        ╱ Test?  ╲          │  Start  │
└─────────────┘        ╲        ╱          ╰─────────╯
                        ╲──────╱

Flow Control:

Standard Flow:      Branch:           Parallel:         Merge:
     │              │   │   │         ┌────┬────┐         │
     ▼              ▼   ▼   ▼         │    │    │      ───┴───
                                      ▼    ▼    ▼         │

7 Core Patterns:

Workflow: Select pattern → Build with components → Validate (validate_flowchart.sh) → Document

4. 📋 RULES

Mode 1: Document Quality

✅ ALWAYS

1. ALWAYS validate filename conventions (snake_case, preserve README.md/SKILL.md)
2. ALWAYS detect document type first (applies correct enforcement level)
3. ALWAYS verify frontmatter for SKILL.md and Command types
4. NEVER add TOC (only allowed in README files)
5. ALWAYS ask about llms.txt generation (never auto-generate)
6. ALWAYS apply safe auto-fixes (H2 case, separators, filenames)
7. ALWAYS validate before completion (structure + content + style)
8. ALWAYS provide metrics (before/after counts from script output)

❌ NEVER

1. NEVER modify spec files during active development (loose enforcement)
2. NEVER delete original content without approval
3. NEVER block for safe violations (only block: missing frontmatter, wrong order)
4. NEVER generate llms.txt without asking
5. NEVER apply wrong enforcement level

⚠️ ESCALATE IF

1. Document type ambiguous
2. Critical violations detected
3. Major restructuring needed
4. Style guide missing
5. Conflicts with user intent

Mode 2: OpenCode Component Creation

Skills

✅ ALWAYS

1. ALWAYS start with concrete examples (validate understanding)
2. ALWAYS run init_skill.py (proper scaffolding)
3. ALWAYS identify bundled resources (scripts/references/assets)
4. ALWAYS use third-person ("Use when..." not "You should use...")
5. ALWAYS keep SKILL.md <5k words (move details to references/)
6. ALWAYS delete unused examples (keep lean)
7. ALWAYS validate before packaging
8. ALWAYS recommend final review (run extract_structure.py)

❌ NEVER

1. NEVER use second-person (imperative/infinitive only)
2. NEVER duplicate SKILL.md/references/ (progressive disclosure)
3. NEVER create without examples
4. NEVER skip validation
5. NEVER include excessive detail (SKILL.md is orchestrator)
6. NEVER use vague descriptions

⚠️ ESCALATE IF

1. Skill purpose unclear
2. No concrete examples
3. Validation fails repeatedly
4. Unsupported features
5. User input required (brand assets, API docs)

Agents

✅ ALWAYS

1. ALWAYS load agent_template.md first (template-first workflow)
2. ALWAYS validate frontmatter (name, mode, temperature, tools, permission)
3. ALWAYS include CORE WORKFLOW section (numbered steps)
4. ALWAYS include ANTI-PATTERNS section (what NOT to do)
5. ALWAYS set explicit tool permissions (true/false for each tool)
6. ALWAYS test with real examples before deployment

❌ NEVER

1. NEVER create agents without @write agent (bypasses quality gates)
2. NEVER skip frontmatter validation (causes discovery failures)
3. NEVER use vague tool permissions (be explicit: true or false)
4. NEVER omit anti-patterns (agents need clear boundaries)

⚠️ ESCALATE IF

1. Agent purpose overlaps with existing agent
2. Tool permissions unclear
3. Behavioral rules conflict with AGENTS.md

Commands

✅ ALWAYS

1. ALWAYS load command_template.md first (template-first workflow)
2. ALWAYS define clear triggers (what invokes the command)
3. ALWAYS include usage examples (copy-paste ready)
4. ALWAYS validate command name (lowercase, colon-separated)

❌ NEVER

1. NEVER create commands without frontmatter (required for discovery)
2. NEVER use ambiguous triggers (must be unique)
3. NEVER skip testing (commands must work on first invocation)

⚠️ ESCALATE IF

1. Command conflicts with existing command
2. Trigger phrase is ambiguous
3. Command requires special permissions

Mode 3: Flowchart Creation

✅ ALWAYS

1. ALWAYS use consistent box styles (single-line process, rounded terminals, diamond decisions)
2. ALWAYS label all decision branches (Yes/No or specific outcomes)
3. ALWAYS align elements (no diagonal lines, consistent spacing)
4. ALWAYS show complete paths (every box has entry/exit)
5. ALWAYS validate readability

❌ NEVER

1. NEVER create ambiguous arrow connections
2. NEVER leave decision outcomes unlabeled
3. NEVER exceed 40 boxes (break into sub-workflows)
4. NEVER mix box styles inconsistently
5. NEVER skip spacing and alignment

⚠️ ESCALATE IF

1. Process exceeds ~40 boxes
2. Interactive/exportable format needed
3. Collaborative editing required
4. Pattern unclear

Mode 4: Install Guide Creation

✅ ALWAYS

1. ALWAYS include AI-first install prompt at the top
2. ALWAYS use phase validation checkpoints (phase_N_complete pattern)
3. ALWAYS provide platform-specific configurations (OpenCode, Claude Code, Claude Desktop)
4. ALWAYS include troubleshooting section with Error → Cause → Fix format
5. ALWAYS verify commands are copy-paste ready

❌ NEVER

1. NEVER skip validation checkpoints (each phase must validate)
2. NEVER assume prerequisites (always list and verify)
3. NEVER mix platform instructions (separate clearly)
4. NEVER use relative paths in command examples

⚠️ ESCALATE IF

1. Multi-platform complexity requires testing
2. External dependencies unavailable
3. Installation requires special permissions

Emoji Usage Rules

Body Text: ✅ Status indicators (✅ ❌ ⚠️), priority markers (🔴 🟡 🔵), visual indicators (📊 🔍 ⚡) - only when enhancing clarity.

H3 Semantic Exception: Emojis ✅ ❌ ⚠️ REQUIRED on H3 in RULES sections for functional signaling.

H2 Emoji Enforcement by Document Type

CRITICAL: For template-based documents (SKILL, README, asset, reference), missing H2 emojis are BLOCKING errors. The extract_structure.py script will return severity: 'error' for these violations.

Prevention: Always COPY headers from templates. Never reconstruct from memory.

5. 🏆 SUCCESS CRITERIA

Document Quality Index (DQI)

The extract_structure.py script computes a DQI (0-100) based on measurable attributes:

Quality Bands:

Example DQI Output (from extract_structure.py):

{
  "dqi": {
    "total": 96,
    "band": "excellent",
    "components": {
      "structure": 40,
      "content": 26,
      "style": 30
    }
  },
  "checklist": { "passed": 12, "failed": 0, "skipped": 2 },
  "documentType": "SKILL"
}

Completion Checklists

Document Quality Complete:

• ✅ extract_structure.py executed, JSON parsed
• ✅ Document type detected, checklist reviewed
• ✅ Evaluation questions answered, recommendations generated
• ✅ All critical issues addressed

Skill Creation Complete:

• ✅ YAML frontmatter with name + description (third-person, specific)
• ✅ SKILL.md under 5k words, bundled resources organized
• ✅ Unused examples deleted, passes package_skill.py
• ✅ Final AI review completed, tested on real examples

Agent Creation Complete:

• ✅ YAML frontmatter with name, mode, temperature, tools, permission
• ✅ Tool permissions explicitly set (true/false for each)
• ✅ CORE WORKFLOW section with numbered steps
• ✅ ANTI-PATTERNS section with clear boundaries
• ✅ RELATED RESOURCES section with links
• ✅ Tested with real examples

Command Creation Complete:

• ✅ YAML frontmatter with name, description, triggers
• ✅ Clear usage examples (copy-paste ready)
• ✅ Execution logic defined
• ✅ Added to command registry
• ✅ Tested invocation works

Flowchart Complete:

• ✅ All paths clear, decisions labeled, parallel blocks resolve
• ✅ Spacing consistent, understandable without explanation
• ✅ Size limits: ≤40 boxes, ≤8 depth levels, ≤200 lines

Install Guide Complete:

• ✅ AI-first prompt included, copy-paste ready
• ✅ All 5 phases have validation checkpoints
• ✅ Platform configurations provided (at least OpenCode)
• ✅ Troubleshooting covers common errors
• ✅ Commands tested and working

Document-Type Gates

6. 🔌 INTEGRATION POINTS

Framework Integration

This skill operates within the behavioral framework defined in AGENTS.md.

Key integrations:

• Gate 2: Skill routing via skill_advisor.py
• Tool Routing: Per AGENTS.md Section 6 decision tree
• Memory: Context preserved via Spec Kit Memory MCP

Scripts

Tool Usage

Related Skills

Workflow Integration

Skill Creation → Document Quality:

1. Initialize (init_skill.py)
2. Edit SKILL.md and resources
3. Package (package_skill.py)
4. Quality validation (extract_structure.py)
5. Iterate if needed

7. 📚 EXTERNAL RESOURCES

• llms.txt specification: https://llmstxt.org/
• Context7 (external docs benchmark): https://context7.ai/
• Anthropic documentation: https://docs.anthropic.com/
• CommonMark specification: https://spec.commonmark.org/

8. 🔗 RELATED RESOURCES

For Document Quality

1. Read Sections 3-6 (When/How/Rules/Success)
2. Navigate: workflows.md for execution modes
3. Run enforcement, optimization, or validation as needed

For Skill Creation

1. Read Sections 3-6 (When/How/Rules/Success)
2. Navigate: skill_creation.md for workflow
3. Use Scripts: init_skill.py → edit → package_skill.py
4. Validate: Run Document Quality validation on SKILL.md

Quick Reference

Need fast navigation? See quick_reference.md

Remember: This skill operates in four modes - Document Quality, Skill Creation, Flowchart Creation, and Install Guide Creation. All modes integrate seamlessly for creating and validating high-quality documentation and skills.