Ml Architecture Diagram

Create accurate, publication-ready, editable machine-learning and deep-learning architecture diagrams from model code, configuration files, model summaries, graph exports, or written specifications. Use for neural-network figures, architecture schematics, model block diagrams, training/inference diagrams, paper figures, and editable SVG/draw.io/PPTX outputs. Recover the real computation graph first, compress unimportant implementation detail, emphasize the scientific contribution, and never invent modules for aesthetics.

5 starsAI, LLMs & Agents

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add Ztsdut/ml-architecture-diagram-skill --skill ml-architecture-diagram

Manual install

git clone --depth 1 https://github.com/Ztsdut/ml-architecture-diagram-skill.git /tmp/ml-architecture-diagram-skill
mkdir -p ~/.claude/skills
cp -r /tmp/ml-architecture-diagram-skill/. ~/.claude/skills/ml-architecture-diagram

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# ML Architecture Diagram

Create architecture figures that are both **faithful to the model** and **useful to a scientific reader**.

The job has two distinct stages:

1. recover and normalize the architecture into the repository's Architecture IR;
2. render or redesign that IR into a publication-ready figure, choosing deterministic vector rendering, AI-assisted rendering, or both.

Do not jump directly from a class listing to a pretty diagram.

## 1. Determine the requested figure

Infer the user's goal from the request and supplied files.

Common modes:

- **paper-overview** — compact model overview for a manuscript;
- **paper-detail** — overall architecture plus one or more custom-block details;
- **training-pipeline** — architecture plus loss/objective/training-only paths;
- **inference-pipeline** — runtime data flow only;
- **presentation** — more explanatory spacing and labels;
- **exhaustive-debug** — more implementation detail for engineering review.

Default to `paper-overview` unless the request clearly implies another mode.

## 2. Recover the true computation graph

Use the executed inference path as the source of truth.

### General rules

Identify:

- inputs and their semantic meaning;
- preprocessing and embeddings;
- serial and parallel branches;
- repeated blocks;
- residual/skip routes;
- merge operations such as add, concatenate, gating, multiplication, routing, or cross-attention;

Read the full skill on GitHub

Details

Skill name: ml-architecture-diagram
Repository: Ztsdut/ml-architecture-diagram-skill
License: MIT
Repo updated: 12 August 2026

FAQ

What does the Ml Architecture Diagram skill do?

How do I install the Ml Architecture Diagram skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add Ztsdut/ml-architecture-diagram-skill --skill ml-architecture-diagram. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Ml Architecture Diagram popular?

It lives in Ztsdut/ml-architecture-diagram-skill, which has 5 GitHub stars and 0 forks.

Related skills

No AI Tone

mckenizemanfull5678-ctrl/no-ai-tone

中文长文写作五步管线：把杂乱素材整理成一篇读起来「不像 AI 写的」文章。当用户要写公众号 / 知乎 / 博客等中文长文，要把素材·笔记·链接整理成稿，抱怨 AI 写的东西有「AI 腔 / AI 味」，或想给一篇已有文章去 AI 腔、提炼金句、规划配图时使用。主线五步写作 + 配图落地：选题研判 → 内容线设计 → 大纲打磨 → 风格化初稿 → 双重精修（去 AI 腔 + 金句提炼） → 配图落地（信息图为主、配图也去 AI 味）。核心心法是「人决策、AI 执行」——每一步产出后停下来等用户拍板，绝不一口气跑完。

AI, LLMs & Agents

Industrial Waste Heat Recovery

dungnotnull/industrial-waste-heat-recovery-agent-skill

Industrial Waste-Heat Recovery System Design — Production-grade evidence-backed analysis harness with flexible agent architecture, skill registry, and modular sub-agents. Use for: waste-heat recovery design, heat exchanger selection, ORC systems, pinch analysis, energy efficiency assessment, industrial thermal system optimization. Triggers on: "waste heat", "heat recovery", "ORC", "pinch analysis", "heat exchanger", "energy efficiency", "thermal optimization", "industrial heat", "steam system", "condensate recovery".

AI, LLMs & Agents

Oeconomia

vincitamore/oeconomia

Delegation economy for a session whose own context window is the scarce resource. REFLEX: name the model wherever the dispatch primitive accepts one (an unset model can inherit the session model and silently spend the budget you protect); delegate bulk early (multi-file reads, protocol-driven review sweeps, build/test runs, scripted edits) so freight never lands in the compounding window; keep judgment and every required orientation read; floor: one known-file read, one grep, a two-line edit, or any read near session end stays home; a same-provider subagent relieves the window, never the accou

AI, LLMs & Agents

Human Maintainability

hmr-BH/HMAP

Assess a programming project's HUMAN maintainability (HMAP): 0-100 score, nine-tier verdict, per-dimension scores, and file:line evidence — built to catch vibe-coding / AI-generated slop, including semantic incoherence (name/contract/concept/vocabulary drift). Triggers: assess maintainability / rate this project / code quality score / readability score / AI slop detection / vibe coding check / semantic consistency check / HMAP / human maintainability / 评估项目可维护性 / 给项目打分 / 代码质量评估 / 可读性评分 / AI屎山检测 / vibe coding 体检 / 是否适合人类维护 / 语义一致性. Anti-trigger: the user asks to fix, refactor, optimize, or writ

AI, LLMs & Agents

Omi

CzzzzzzJ/omi

Omi3.0 小红书图文笔记生产线。输入内容素材 + IP形象，自动产出分镜脚本JSON + 图片生成指令，适用于 Claude Code、OpenClaw、Codex、扣子等任何具备本地图片生成能力的 Agent 平台。触发词：omi、小红书笔记生成、omi笔记

AI, LLMs & Agents

Goodcode

SpinaBuilds/goodcode

Host-agnostic exhaustive multi-agent orchestration for any coding agent that can spawn subagents (Claude Code, Codex, Cursor, OpenCode, OpenHands, …). Pipeline: recon → cheap baseline → a-priori specialist roles cast by routing → parallel fan-out → adversarial verification → loop-until-dry → dedup/score/synthesize. Optimizes for the most exhaustive, correct answer — not the fastest or cheapest. Use for audits, deep reviews, multi-source research, design exploration, large migrations, or any substantial task where completeness beats latency. Triggers: '/goodcode', 'goodcode', 'orchestrate exhau

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.