doc-generator

name	doc-generator
description	Use this skill when the user wants to generate, update, or improve usage documentation for a project, file, component, SDK, module, or function based on source code.

Doc Generator Skill

Purpose

This skill analyzes codebases, files, and symbols, then generates practical usage documentation.

The goal is not only to explain code, but to produce documentation that helps users understand:

• what this code/module does
• when it should be used
• how to use it correctly
• what constraints or risks exist
• what common misuse patterns should be avoided

This skill is intended for:

• component libraries
• SDKs
• utility libraries
• infrastructure libraries
• business shared modules
• core project modules

Supported Targets

This skill supports the following target scopes:

1. project-level

   • entire repository
   • package
   • component library
   • SDK project

2. file-level

   • single file
   • module file
   • entry file
   • service / manager / composable / hook file

3. symbol-level

   • function
   • method
   • class
   • component
   • hook / composable
   • exported constant / config factory

Core Principles

1. Documentation must be usage-oriented, not just implementation-oriented.
2. Prefer explaining how to use, when to use, and what to watch out for.
3. Prefer minimal runnable examples over abstract explanation.
4. Do not invent behavior that cannot be supported by code evidence.
5. Clearly mark uncertainty when behavior cannot be fully confirmed from source.
6. Distinguish between:
   • confirmed by code
   • high-confidence inference
   • requires verification
7. Generated docs must be readable by engineers who did not write the code.
8. When existing docs already exist, prefer update / improve / align, not blindly overwrite.

Default Workflow

When invoked, follow this workflow:

1. Determine target scope:

   • project
   • file
   • symbol

2. Determine documentation intent:

   • readme
   • usage-guide
   • api-reference
   • module-guide
   • best-practices
   • faq
   • migration-guide
   • update-existing-doc

3. Inspect available evidence:

   • folder structure
   • exports
   • type definitions
   • comments
   • function signatures
   • call relationships
   • example code
   • tests
   • existing documentation
   • package metadata
   • config files

4. Build understanding:

   • responsibility
   • public surface
   • dependencies
   • usage flow
   • constraints
   • edge cases
   • risks
   • example patterns

5. Generate documentation in the requested format.

6. Run quality checks before finalizing.

Required Behavior

Always do

• explain the target’s responsibility in plain language
• identify public-facing APIs or exposed usage surfaces
• describe suitable scenarios and, if possible, unsuitable scenarios
• include at least one usage example when enough evidence exists
• include caveats / limitations / risk notes when relevant
• explicitly state uncertainty when evidence is insufficient

Never do

• fabricate business context
• fabricate runtime behavior
• claim compatibility not supported by evidence
• describe internal assumptions as facts
• output vague filler without concrete value
• rewrite code as documentation without interpretation

Documentation Style

Documentation should be:

• clear
• structured
• practical
• concise but sufficiently informative
• focused on real usage

Prefer:

• direct statements
• scenario-based explanation
• minimal examples
• explicit constraints
• actionable warnings

Avoid:

• empty praise
• generic summaries
• repeating code names without interpretation
• over-explaining obvious syntax

Uncertainty Policy

When the code does not provide enough evidence:

• mark the statement as Inference
• or mark as Needs Verification

Examples:

• Confirmed: visible directly from exports, types, or implementation
• Inference: strongly suggested by naming, call chain, or surrounding structure
• Needs Verification: depends on runtime, external system, hidden contract, or business rules

Output Preference

Unless the user specifies otherwise:

• project-level → produce README or usage guide
• file-level → produce module guide
• symbol-level → produce API reference

Reference Files

Read and follow:

• references/overview.md
• references/input-schema.md
• references/output-rules.md
• references/analysis-rules.md
• references/doc-templates.md
• references/quality-checklist.md

Success Criteria

This skill succeeds when the produced documentation helps a new engineer answer:

• What is this for?
• When should I use it?
• How do I use it?
• What do I need before using it?
• What can go wrong?
• What is the safest recommended usage pattern?