Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    draw architecture

    1

    Generates publication-quality AI and software architecture diagrams in TikZ, SVG, and JSON.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    draw architecture

    draw architecture

    Example session with this skill installed

    Generate a professional diagram for a RAG system with a vector database, an LLM, and a retrieval reranker. Export as SVG and TikZ.

    • Read your context and instructions
    • Compiled the draw architecture

    I have generated the RAG system architecture diagram. I extracted the components (User, Query, Vector DB, Reranker, and LLM) and established a sequential data flow with a feedback loop for the reranking stage.

    RAG System Architecture

    ComponentFunctionInterface
    User QueryEntry PointString Input
    Vector DBSemantic RetrievalTop-K Embeddings
    RerankerPrecision FilteringScored Documents
    LLMGenerationFinal Response
    {
      "nodes": [
        {"id": "user", "label": "User", "type": "actor", "x": 0, "y": 0},
        {"id": "vdb", "label": "Vector DB", "type": "storage", "x": 2, "y": 0},
        {"id": "rerank", "label": "Reranker", "type": "process", "x": 4, "y": 0},
        {"id": "llm", "label": "LLM", "type": "model", "x": 6, "y": 0}
      ],
      "edges": [
        {"from": "user", "to": "vdb", "label": "query"},
        {"from": "vdb", "to": "rerank", "label": "candidates"},
        {"from": "rerank", "to": "llm", "label": "context"},
        {"from": "llm", "to": "user", "label": "response", "kind": "dashed"}
      ]
    }
    

    Figure: A standard RAG pipeline showing retrieval flow from Vector DB through a Reranker to the LLM, with a dashed response path to the user.
    Source: output/rag_arch.svg, output/rag_arch.tex, output/rag_arch.json.
    Verified: Semantic checks passed; no overlapping routes; label positions optimized.
    Assumptions: Used standard cosine similarity for VDB; assumed synchronous processing.

    Next steps

    • Run python $SKILL_ROOT/scripts/archdiagram.py render rag_arch.json --theme paper to update typography.
    • Open rag_arch.tex in Overleaf for final LaTeX paper integration.
    • Add via waypoints to the JSON if you need to route the response line around the Reranker node.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Architecture diagrams in technical docs and papers are often imprecise, messy, or impossible to edit once rendered. This skill generates exact, publication-quality schematics for AI systems, neural networks, and software services from code or descriptions.

    What it does

    • Generate schematics for RAG, multi-agent systems, and service topologies using a custom SVG/TikZ renderer.
    • Visualize neural architectures including Transformers, CNNs, and GNNs with precise tensor shapes and data flow.
    • Extract evidence from PyTorch code, papers, or configs to build diagrams grounded in actual implementation.
    • Search a figure atlas containing over 100 TikZ master templates for vision, audio, robotics, and LLM components.
    • Export editable formats including raw TikZ for LaTeX, structured JSON scenes, and production-ready SVG/PDF.

    How it works

    1. Analyze evidence by inspecting provided source code or text to identify components, interfaces, and tensor dimensions.
    2. Select a route between the JSON-to-SVG renderer for systems or native TikZ templates for mathematical neural layers.
    3. Build the source using archdiagram.py to validate layout, handle edge routing, and manage group boundaries.
    4. Run semantic review to verify arrow directions, resolve label collisions, and confirm residual connection origins.

    Frameworks & tools

    Built on a Python 3.10+ stdlib renderer. Supports TikZ/LaTeX for math-heavy figures and integrates with Inkscape for PDF/PNG exports.

    Why this beats prompting it yourself

    Standard LLMs struggle with spatial reasoning, leading to overlapping nodes and "hallucinated" connections in Mermaid or Graphviz. This skill uses a specialized geometric renderer and a vetted atlas of 100+ professional templates to ensure mathematical and architectural accuracy.

    Use cases

    • Research paper figures that require monochrome LaTeX typography and TikZ source.
    • System documentation for multi-agent workflows and RAG pipelines.
    • Code-to-diagram conversion to visualize existing PyTorch or backend service structures.
    • Technical presentations needing editable, high-resolution vector schematics.

    Known limitations

    Does not train models or infer missing code logic. High-fidelity PDF/PNG export requires a local Inkscape installation. TikZ compilation requires a TeX engine.

    Requires Python 3.10+. Native TikZ compilation needs a local TeX engine. PNG/PDF exports require Inkscape. Cannot perfectly infer architecture from obfuscated or incomplete code.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Requires

    Python
    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    i will update every 6 months

    Listed29 days ago
    Updated8 days ago

    What's inside

    Frequently Asked Questions

    Unlike standard LLMs that often produce overlapping nodes or broken logic, this skill uses a specialized geometric renderer and a vetted atlas of 100+ templates to ensure spatial accuracy and valid connection routing.

    You will receive the 'draw architecture' skill configuration, access to the library of TikZ master templates, and the custom Python-based renderer used to generate precise SVG and JSON outputs.

    Yes, the skill can analyze PyTorch code or configuration files to extract components and tensor shapes, ensuring your diagrams are grounded in your actual technical implementation.

    The skill exports raw TikZ for LaTeX integration, structured JSON for custom pipelines, and production-ready SVG; high-fidelity PDF and PNG exports are supported if you have Inkscape installed locally.

    While the skill generates the code and layout, you will need a TeX engine (like TeX Live or Overleaf) to compile TikZ files and a Python 3.10+ environment to run the underlying renderer.

    New in Agents & Orchestration