meta-diagram-triangulation

Generates PlantUML and draw.io architecture diagrams from a codebase scan in parallel.

6.8k|540|Updated May 6, 2026
One-click install
npx skills add https://github.com/opensquilla/opensquilla --skill meta-diagram-triangulation
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: meta-diagram-triangulation
Source: https://github.com/opensquilla/opensquilla/tree/main/src/opensquilla/skills/exp/meta-diagram-triangulation
Command: npx skills add https://github.com/opensquilla/opensquilla --skill meta-diagram-triangulation

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Writing architecture documentation for RFCs or onboarding requires both a text-friendly diagram for version control and an editable diagram for presentations, and producing both manually from a codebase is slow and inconsistent.

Core Features & Use Cases

  • Automated Codebase Scanning: Extracts top-level modules, inter-module dependencies, hotspot files, and public API surface from a target path.
  • Diagram Kind Classification: An LLM classification step picks the most informative diagram type (class, sequence, component, deploy, or flow) based on the scan results.
  • Parallel Dual Rendering: Generates both a PlantUML source file and a draw.io XML file from the same scan, then composes them into a single architecture docx and persists a snapshot to memory.
  • Use Case: When drafting an RFC for a new service, invoke the skill on the service's source directory to receive an arch.puml for git review, an arch.drawio for stakeholder editing, and an arch.docx tying them together.

Quick Start

Run diagram triangulation on the src/opensquilla directory to produce PlantUML and draw.io architecture diagrams plus a composed docx.

Frequently Asked Questions about meta-diagram-triangulation

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate architecture diagrams from an existing codebase?

Invoke the skill with a target path in your message, and it scans modules, dependencies, hotspots, and public surfaces, then renders the result as both PlantUML source and draw.io XML. The outputs are written to a diagrams directory in your workspace.

What diagram types can be generated from a code scan?

The classification step chooses one of five kinds: class, sequence, component, deploy, or flow. The choice is based on whether the scan shows dominant data structures, call flows, module boxes, infrastructure layout, or data pipelines.

PlantUML vs draw.io for architecture documentation, which does this produce?

It produces both in parallel from the same scan. PlantUML source suits text-based git review, while the draw.io XML file can be opened and rearranged in the draw.io editor for presentations.

What happens if one diagram render step fails?

The compose step still produces the docx referencing whichever render succeeded. You can manually re-run the failed render through the sub-agent skill using the same scan output as input.

Does the skill keep a record of generated architecture diagrams?

Yes, a final persist step appends the invocation, chosen diagram kind, and output file paths to memory/architecture-snapshots.md via the memory_save tool, creating an audit trail of diagram snapshots.