name: agent-ops-create-technical-docs description: "Create focused, specific technical documentation for codebase sections. Analyzes code, identifies topics, presents options before writing. Supports code blocks with line numbers." license: MIT compatibility: [opencode, claude, cursor]
metadata: category: docs related: [agent-ops-interview, agent-ops-planning, agent-ops-context-map, agent-ops-state, agent-ops-docs]
Technical Documentation Creation
Purpose
Create focused, specific technical documentation for codebase sections. Unlike general docs (README, CHANGELOG), this skill produces deep-dive articles on components, patterns, or subsystems.
When to Use
- Documenting a complex module or subsystem
- Creating tutorials for internal APIs
- Writing architectural explainers
- Onboarding documentation for new developers
- Self-documenting AgentOps skills and workflows
Core Workflow
Analyze → Discover Topics → Present Options → Write → Review
Phase 1: Analysis
Input Types
| Input | Scope | Use Case |
|---|---|---|
| Entire codebase | Broad discovery | Initial documentation audit |
| Folder/module | Scoped | Document a subsystem |
| Single file | Detailed | Deep dive on one component |
| Code snippet | Targeted | Explain specific pattern |
Analysis Procedure
- Read context map (
.agent/map.md) if available - Scan target scope:
- Entry points and exports
- Public API surface
- Dependencies (imports/requires)
- Configuration points
- Check existing docs:
.agent/docs/— agent-generated docsdocs/— project docsREADME.md— top-level overview
- Output: List of components with documentation potential
Phase 2: Topic Discovery
Topic Types
| Type | Description | Best For |
|---|---|---|
| How-To | Step-by-step task completion | Practical tasks |
| Explainer | Concept/design rationale | Understanding "why" |
| Reference | API/config/schema details | Quick lookup |
| Tutorial | Learning path with examples | Onboarding |
| Deep Dive | Internal implementation details | Advanced users |
Topic Sizing Guidelines
Good topics are:
- Readable in 5-15 minutes
- Self-contained (minimal external dependencies)
- Actionable (reader can do something) or educational (reader understands something)
Split large topics into series if needed.
Discovery Output
Generate topic candidates with:
- Title (action-oriented or descriptive)
- Type (How-To, Explainer, etc.)
- Target audience (beginner, intermediate, advanced)
- Estimated length (short: <500 words, medium: 500-1500, long: >1500)
- Overlap check (similar existing docs)
Phase 3: Interactive Selection
Use agent-ops-interview to present topics before writing.
Topic Presentation Format
📚 **Documentation Topics for `src/bundle/installer.py`**
I identified 5 potential documentation topics:
1. **[How-To]** Installing bundles with category filtering
- Audience: End users | Length: Medium
2. **[Explainer]** How the conflict resolution algorithm works
- Audience: Contributors | Length: Long
3. **[Reference]** BundleInstaller public API
- Audience: Developers | Length: Medium
4. **[Deep Dive]** Frontmatter parsing and category discovery
- Audience: Advanced | Length: Long
5. **[Tutorial]** Creating a custom bundle category
- Audience: Beginners | Length: Medium
Which topics should I write? (comma-separated numbers, 'all', or 'suggest')
Selection Options
| Response | Action |
|---|---|
1,3,5 | Write selected topics |
all | Write all topics |
suggest | Agent recommends based on gaps |
none | Skip documentation |
refine 2 | Discuss topic 2 scope before deciding |
Phase 4: Writing
Code Block Standards
Always include:
- File path
- Line numbers (when referencing specific code)
- Syntax highlighting
Format:
```python title="src/bundle/installer.py" linenums="42"
def load_category_manifest(bundle_path: Path) -> CategoryManifest:
"""Load category definitions and build dynamic mappings."""
manifest_path = bundle_path / "bundle-categories.yaml"
if not manifest_path.exists():
return CategoryManifest(categories={}, presets={})
# ...
```
Code Extraction Rules
- Show relevant context — not entire files
- Highlight key lines — use comments or annotations
- Link to source — provide file path for full code
- Keep snippets runnable — when possible, show complete examples
Structure by Type
How-To Structure
# How to [Action]
## Prerequisites
- Requirement 1
- Requirement 2
## Steps
### 1. [First Step]
[Explanation + code]
### 2. [Second Step]
[Explanation + code]
## Result
[What the user should see/have]
## Troubleshooting
| Problem | Solution |
|---------|----------|
| Error X | Fix Y |
Explainer Structure
# Understanding [Concept]
## What It Is
[One paragraph definition]
## Why It Exists
[Problem it solves, context]
## How It Works
[Detailed explanation with diagrams/code]
### [Subcomponent 1]
[Details]
### [Subcomponent 2]
[Details]
## Trade-offs
| Approach | Pros | Cons |
|----------|------|------|
## Related Concepts
- [Link to related doc 1]
- [Link to related doc 2]
Reference Structure
# [Component] Reference
## Overview
[Brief description]
## API
### `function_name(param1, param2)`
**Parameters:**
| Name | Type | Required | Description |
|------|------|----------|-------------|
**Returns:** `ReturnType` — Description
**Raises:**
- `ErrorType` — When condition
**Example:**
```language
// Usage example
Configuration
| Option | Type | Default | Description |
|---|
See Also
- [Related reference]
#### Tutorial Structure
```markdown
# Tutorial: [Learning Goal]
## What You'll Learn
- Skill 1
- Skill 2
## Prerequisites
- Knowledge required
- Tools needed
## Part 1: [Foundation]
[Explanation + hands-on]
## Part 2: [Building]
[Explanation + hands-on]
## Part 3: [Completing]
[Explanation + hands-on]
## Summary
[Recap of skills learned]
## Next Steps
- [Advanced topic link]
- [Related tutorial link]
Deep Dive Structure
# Deep Dive: [Internal Component]
> **Audience**: Contributors and advanced users
## Architecture Overview
[High-level diagram/description]
## Implementation Details
### [Subsystem 1]
[Code walkthrough with annotations]
### [Subsystem 2]
[Code walkthrough with annotations]
## Key Decisions
| Decision | Rationale | Alternatives Considered |
|----------|-----------|------------------------|
## Extension Points
[How to customize/extend]
## Known Limitations
- Limitation 1
- Limitation 2
## Future Considerations
- Potential improvement 1
Phase 5: Review
Pre-Publish Checklist
- Accuracy: Code snippets match current source
- Completeness: All sections filled, no TODOs
- Links: No broken internal links
- Code validity: Snippets parse/compile
- No secrets: No API keys, passwords, internal URLs
- Audience fit: Complexity matches stated audience
Overlap Check
Before finalizing, verify no duplication:
- Search existing docs for similar content
- If overlap found, offer to update existing instead of creating new
Output Locations
| Content Type | Location |
|---|---|
| Agent-generated docs | .agent/docs/ |
| Project docs | docs/ |
| MkDocs-ready | docs/ with frontmatter |
MkDocs Frontmatter
When outputting to docs/ for MkDocs:
---
title: "Document Title"
description: "Brief description for SEO/preview"
date: 2026-01-21
tags:
- bundle
- installation
---
Self-Documentation Mode
This skill can document AgentOps itself:
Suggested topics:
- Skill system architecture
- Issue management workflow
- Bundle installer mechanics
- State file conventions
- Interview protocol patterns
To self-document:
Analyze .github/skills/ and create documentation for the AgentOps skill system
Integration Points
| Skill | Integration |
|---|---|
agent-ops-context-map | Use map.md for codebase overview |
agent-ops-docs | Coordinate README/CHANGELOG updates |
agent-ops-mkdocs | Output compatible with MkDocs nav |
agent-ops-interview | Topic selection dialog |
Anti-Patterns
- ❌ Writing before presenting topic options
- ❌ Entire file dumps instead of focused snippets
- ❌ Code without line numbers or file paths
- ❌ Generic content that doesn't reference actual code
- ❌ Creating docs that duplicate existing content
- ❌ Skipping the audience/complexity assessment
- ❌ Outdated code snippets that don't match source