Nanosaur2-670M
well9472 · Kit
Makes illustrations from text prompts with a small 670M-parameter diffusion model, run through ComfyUI. Built as a low-budget research model for people who wan…
New
Overview
Lex from SigmaStudioz’s six skills for coding agents. Three shape engineering work: domain-driven changes, timeboxed spikes and outcome reviews. Three are visual: interactive diagrams and films, video analysis and bundled React effects.
SigmaStudioz Agent Skills is an open-source project on GitHub in the Developer category, listed by Mindket. It is free. This is version 2, last updated on . Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account, as a zip.
At a glance
- Runs in
- Claude Code: the skills tool, or the plugin marketplace with claude plugin commands
- Codex: the skills tool, or a local clone registered with codex plugin marketplace add
- Cursor: the skills tool, or its native skill folder, plus a global rule for automatic use
- Any other agent: the skills tool, which asks you to choose your coding agent
- Start with
npx skills add sigmastudioz/agent-skills --skill domain-driven-design
Platform
Claude Code
Codex
Cursor
About
What it is
SigmaStudioz Agent Skills is Lex from SigmaStudioz’s collection of six skills for coding agents, each in its own plugin with its own version. Three shape engineering work. Principal Driven Domain Design keeps code with the domain that owns it and traces gaps before a handoff. Engineering Spikes runs timeboxed research and writes two documents, one for stakeholders and one for engineers. Does It Make Sense? checks a finished build against its original goal. Three are visual. Visualizer builds interactive diagrams and films, Video Analysis reads a clip through timestamped frames, and UI Foundation Reuse brings bundled React effects.
- You install the skills you want
- Claude loads one when your request fits
- Claude follows its steps on your project
- You get code, a document or a short report
Who it suits
Developers who give a coding agent real features across an API, a worker or a monorepo, and tech leads who need a spike written for two audiences. Designers and developers who build interactive explainers or short films will use the visual three. The domain-driven-design README says small fixes stay small and read-only reviews stay read-only.
| Who | Skill | Example |
|---|---|---|
| Developers on apps with several consumers | domain-driven-design | A developer asks for refunds in an API and a retry worker, and Claude reuses the shared billing code in both. |
| Tech leads | engineering-spikes | A lead asks for a one-day spike on whether the export flow can handle larger reports, and gets disproof criteria, compared options and both documents. |
| Anyone reviewing finished work | does-it-make-sense | A tester asks whether a feature with passing tests still lets the user reach the goal. |
| Explainer and film makers | visualizer, video-analysis, ui-foundation-reuse | A designer asks Claude to chart three months of order counts, with a zero baseline and a data table. |
Where to use it
Claude Code, Codex and Cursor each have a route the maker documents. The install tool also reaches other agents.
- Claude Code: the skills tool, or the plugin marketplace with
claude plugincommands - Codex: the skills tool, or a local clone registered with
codex plugin marketplace add - Cursor: the skills tool, or its native skill folder, plus a global rule for automatic use
- Any other agent: the skills tool, which asks you to choose your coding agent
How to set it up
Get it saves a zip of Mindket’s setup notes in your downloads folder, one file for each route the maker documents. The tool you run installs the skill, and you choose the agent.
| Need | Detail |
|---|---|
| Node.js 22.20 or later | The skills install tool is an npm package, and its README names this version |
| A coding agent | Claude Code, Codex, Cursor or another agent the tool lists |
| Python and FFmpeg | Python 3.10 or later and FFmpeg only for the video sampler and the storyboard renderer. The optional handoff guard needs Python 3 only |
| Accounts and keys | None for the skills. Your agent runs on its own account |
- Open a terminal in your project folder. The skills tool installs into this project by default.
- Run the install command below. The tool asks which agent should get the skill.
- Set up automatic use if you want it. SETUP.md gives this one-time step for domain-driven-design, and the kit’s notes show its three global files.
- Start a new session and ask for a task. Claude loads the skill when your request fits.
npx skills add sigmastudioz/agent-skills --skill domain-driven-design
The other five install the same way with their own names: does-it-make-sense, engineering-spikes, ui-foundation-reuse, visualizer and video-analysis. For Claude Code the maker also documents a plugin route, where the marketplace is called sigmastudioz.
claude plugin marketplace add sigmastudioz/agent-skills claude plugin install domain-driven-design@sigmastudioz --scope user
For Codex, the README clones the repository and registers the clone as a marketplace. It gives no command for installing a single plugin afterwards.
git clone https://github.com/sigmastudioz/agent-skills.git codex plugin marketplace add ./agent-skills
How to use it
Ask for the job a skill is for, and Claude loads the matching one. Only domain-driven-design is written to apply to every coding task. The others start from a request that fits their description.
| Skill | You give it | Claude does |
|---|---|---|
| domain-driven-design | Any coding request: a feature, a fix or a review | Reads the code first, puts changes with the domain that owns them, reuses shared code, traces callers and failure paths, and hands over with evidence |
| engineering-spikes | A decision you cannot make yet, or findings to write up | Frames the question and the timebox, compares options, records evidence and writes one stakeholder document and one engineering document |
| does-it-make-sense | A finished build and its original goal | Explains what exists, traces it back to the goal, gives a verdict for each goal and names the smallest refinements. It edits nothing unless you also authorize fixes |
| visualizer | A subject to explain, some data, or a short film brief | Builds an interactive diagram or simulation, a film, or both, then checks the real export |
| video-analysis | A video or audio file you supply | Inspects timestamped frames and sound, then reports what it saw and what it could not check |
| ui-foundation-reuse | A React project that needs effects or a 3D scene | Looks for an existing effect first, and stages the bundled runtime only when the project has none and the change is authorized |
The visual skills need tools on your computer for some jobs. Video Analysis sampling needs Python 3.10 or later with FFmpeg. The Visualizer renderer needs Python 3.10 or later, Pillow, FFmpeg and two fonts. Claude uses what you already have and says what is missing.
Example
Your billing project has an API and a retry worker, and you need refunds in both. Automatic use is set up, so the request names no skill.
Add refund processing to the API and background job. Keep existing payment behavior working. Follow the current public contracts and run the relevant checks.
Claude looks at the billing domain first, reuses the shared billing code and wires both the API and the worker.
It also follows the retry, authorization and storage paths. That is what the README says the request should lead to, and it asks you to check what the agent loaded, changed and verified.
For newcomers and experts
If you are new. A skill is a folder of instructions that your agent reads when your request fits. You do not need all six. If you write code with an agent, start with domain-driven-design. It changes how the agent works instead of adding a command, so you judge it by the next request. The agent should read before it edits, and it should list what it checked.
For experts. domain-driven-design looks for .principal-ddd/index.md at the repository root. That index links your own tracker or a gap ledger the skill keeps in the same folder. An optional Claude hook, guard_handoff.py, checks handoffs to teammate agents and counts the agent limits you state. Its tests run with one command from a checkout.
python3 -B plugins/domain-driven-design/skills/domain-driven-design/scripts/test_guard_handoff.py
engineering-spikes names its two files with the spike ID, ending in non-technical and technical. By default its output is two basic HTML documents that you open in a browser and paste into Confluence. Publishing needs a destination that you authorize.
Good to know
The maker is open about what its tests do and do not show.
- Evidence: the coding test ran version 1.0.5 on one billing fixture with Claude Code, and an earlier run used Cursor’s command line. No fresh agent test was run for 1.4.0. The README says guidance and scoped tests do not guarantee that every model follows every rule
- Reviews: does-it-make-sense only assesses by default, and a request to review does not authorize edits
- Agents: domain-driven-design does not authorize extra agents, unrelated restructuring or deployment
- Data: PRIVACY.md says there is no publisher backend, account system or telemetry endpoint. Visualizer’s optional sound bridge sends your search terms to Openverse. The skills install tool is a separate project that collects anonymous usage data unless you turn it off
- Licenses: UI Foundation Reuse keeps CC BY notices for bundled head and brain geometry, so keep its notices if you redistribute it
- Maker: Lex from SigmaStudioz, MIT licensed. Not made or endorsed by Anthropic, OpenAI or Cursor
README and files
# SigmaStudioz Agent Skills: setup notes This kit holds setup notes written by Mindket. It does not contain the skills, their helper scripts or the bundled UI runtime. Those install from Lex from SigmaStudioz’s repository: https://github.com/sigmastudioz/agent-skills License: MIT, Copyright (c) 2026 Lex from SigmaStudioz. Full text: https://github.com/sigmastudioz/agent-skills/blob/main/LICENSE. The UI Foundation Reuse plugin also keeps CC BY notices for bundled head and brain geometry, which its README says MIT does not replace. SigmaStudioz Agent Skills is a community project. It is not made or endorsed by Anthropic, OpenAI or Cursor. The pages these notes came from: the README (https://github.com/sigmastudioz/agent-skills/blob/main/README.md), SETUP.md, PRIVACY.md, the three marketplace catalogs, each plugin’s README and SKILL.md under https://github.com/sigmastudioz/agent-skills/tree/main/plugins, the video analysis reference media.md, the visualizer reference storyboards.md and the UI Foundation Reuse reference setup.md. Facts about where Claude Code loads skills and plugins come from Claude Code’s skills page (https://code.claude.com/docs/en/skills). Mindket copied every command and ran none of them. ## What it is Six skills for coding agents, each in its own plugin folder with its own version. | Skill | What it is for | |---|---| | domain-driven-design | Principal Driven Domain Design: code kept with the domain that owns it, shared code reused, integration gaps traced before a handoff | | engineering-spikes | Timeboxed research, evidence, and two separate documents, one for stakeholders and one for engineers | | does-it-make-sense | A finished build traced back to its original goal, with bounded refinements. A review by default, with edits only when you also authorize fixes | | visualizer | Interactive diagrams, data explorers, simulations and visualization videos, with React Three Fiber, shaders and sound | | video-analysis | A video or audio file you supply, read through timestamped frames and a transcript, with the coverage limits stated | | ui-foundation-reuse | A bundled React effects and scene runtime that the agent reuses before it builds its own | Source: the README’s “Available plugins” tables and each SKILL.md’s front matter. ## Files in this kit - setup/skills-cli.md: the `npx skills add` command for each of the six skills - setup/claude-code.md: the plugin route, `claude plugin` commands - setup/codex.md: the clone and `codex plugin marketplace add` route - setup/cursor.md: Cursor’s native skill folder and the global rule - setup/automatic-use.md: the one-time global setup for domain-driven-design, and its optional handoff guard - setup/helpers.md: the optional video sampler, the storyboard renderer and the UI runtime staging helper ## What you need first - A coding agent: Claude Code, Codex, Cursor or another agent the install tool lists. - Node.js 22.20 or later for `npx skills`. The install tool’s README (https://github.com/vercel-labs/skills/blob/main/README.md) says so. - Git, for the Codex route. - Python 3.10 or later and FFmpeg only for the optional video sampler and storyboard renderer, and Node.js for the UI runtime staging helper (setup/helpers.md). The optional handoff guard needs Python 3. - No account, sign-up or API key for the skills. PRIVACY.md says there is no publisher-hosted backend, account system or telemetry endpoint. ## Which route installs what - The skills tool puts one skill folder where your agent reads skills. You choose the agent in its prompt. - The Claude Code plugin route installs a whole plugin from the marketplace named sigmastudioz. - The Codex route registers a local clone as a marketplace. The README says native directory listings are separate from these direct routes, and that installing a skill alone does not configure automatic use across projects. ## Your data The skills are instructions, with optional local helpers. PRIVACY.md says the helpers do not send media or project data to the publisher and that no hook registers the media helpers. It adds that Visualizer’s optional sound bridge sends the search terms and filters you choose to Openverse, then downloads the chosen sound from its source. Whatever your agent reads, and whatever you ask it, goes to your agent’s provider as in any other session. The install tool is a separate project that collects anonymous usage data, and its README names two settings to turn that off. ## Update and remove The README gives no update or removal step for the skills. Claude Code’s skills page says to delete a skill’s folder to remove it, and to disable or uninstall a plugin to remove its skills (setup/claude-code.md). The install tool’s README lists `check`, `update` and `remove` commands.
README.md 5 KB setup/automatic-use.md 4 KB setup/claude-code.md 3 KB setup/codex.md 3 KB setup/cursor.md 2 KB setup/helpers.md 4 KB setup/skills-cli.md 3 KB
What’s new
Updated by Mindket to the current version of this kit.
Reviews
More from Mindket
See all from Mindketwaqf-ocr-hand-written-v1
Waqf-AI · Kit
Reads handwritten Arabic text from images and returns it as text. A fine-tuned PaddleOCR-VL model for people digitizing Arabic handwriting.
New
Wan2.2-TI2V-5B-Diffusers
Wan-AI · Kit
Makes video clips from a text prompt or an image plus text, at up to 720P and 24 frames per second, in the Diffusers format for a single consumer GPU.
New
Similar in Developer
See all in DeveloperWan2.2-S2V-14B
Wan-AI · Kit
Generates cinematic video of a character driven by an audio track, and can also lip-sync existing video. For video creators, animators and researchers with a l…
New
Wan2.2-Animate-14B
Wan-AI · Kit
Animates a character or swaps one into a video by copying a performer's movement and expressions, as one model from the Wan2.2 video family. Use only faces and…
New
d3-turbo
vllm-sr · Kit
Answers choice, yes-or-no, score, box and tracking questions about text, images or video in one call, with a probability for each answer. For people building r…
New