Autonomous Loop

Architecture Documentation Plugin

Three skills for technical documentation: architecture docs with C4 diagrams, technical debrief outlines, and a human writing style toolkit.

Skills

architecture-documentation

Generates architecture documentation from codebases with system diagrams, component analysis, and engineering rationale.

• C4 Model Diagrams (system context, container, component) using PlantUML/Kroki or Eraser syntax
• Component Responsibility Matrix with dependencies, failure modes, recovery strategies
• Configuration rationale, library choices, performance decisions with metrics
• Decision log (ADRs) with context and consequences
• Optional appendices: tech stack summary, API endpoint reference

technical-debrief-outline

Creates structured presentation outlines for technical debriefs. Optimized for mixed audiences (engineers, PMs, executives).

• Problem-solution framing with real examples
• Mermaid diagrams for architecture sections
• Comparison tables (v1 vs v2, POC vs production)
• Presenter callout quotes and audience Q&A prep
• Includes a blank template in references/outline_template.md

human-writing-style

Strips robotic AI patterns from any written content. Produces writing that sounds like a person wrote it.

• Banned word list (delve, leverage, robust, etc.)
• Short sentence / varied rhythm techniques
• Specificity over vagueness — numbers beat adjectives
• Coffee shop test: if it sounds like a LinkedIn post, rewrite it
• Works on docs, outlines, reports, changelogs, anything with words

Installation

/plugin marketplace add dha201/architecture-documentation-plugin
/plugin install architecture-documentation

The marketplace will be registered as dha201-plugins.

Usage

@architecture-documentation Generate architecture documentation for this codebase
@technical-debrief-outline Prepare a debrief for stakeholders
@human-writing-style Rewrite this to sound human

Or let Claude detect when to use them based on your request.

Requirements

• Claude Code 1.0.33 or later

License

MIT