Space Widgets logo

Space Widgets

OrganizationPopular
agent0ai
Space Widgets

Create, patch, inspect, and remove widgets in the current space

Overview

Publisheragent0ai
Repositoryspace-agent
Skill nameSpace Widgets
Stars
1.4K
Forks
323
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by agent0ai on GitHub. Read the source before you install it.

Installation

Install the Space Widgets AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/agent0ai/space-agent.git /tmp/space-agent
mkdir -p .claude/skills
cp -r /tmp/space-agent/app/L0/_all/mod/_core/spaces/ext/skills/space-widgets .claude/skills/agent0ai-space-widgets
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Space Widgets in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Space Widgets on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Space Widgets is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Use this skill for widget work in the current open space

storage

  • ~/spaces//space.yaml = space meta + live layout
  • ~/spaces//widgets/.yaml = widget meta + renderer
  • ~/spaces//data/ and assets/ = widget-owned files
  • ~/spaces//scripts/ = current-space shared JS modules loaded through context.import("scripts/...")

transient

  • Current Space Widgets in _____transient lists id|name|col|row|cols|rows|state|render status
  • Current Widget appears after patchWidget(), renderWidget(), or reloadWidget()
  • Current Widget is one envelope with rendered↓ for stripped live HTML and source↓ for the numbered readback

main helpers

  • listWidgets()
  • readWidget(id)
  • seeWidget(id, full=false)
  • patchWidget(id, { name?, cols?, rows?, col?, row?, metadata?, edits? })
  • renderWidget({ id, name, cols, rows, metadata?, renderer })
  • reloadWidget(id)
  • reposition(), removeWidget(...), removeWidgets(...), removeAllWidgets()
  • rearrangeWidgets(...), toggleWidgets(...), repairLayout(), rearrange(), reload(options?)

catalog and readback

  • Prefer space.current.* for widget work in the current open space
  • Current Space Widgets is the live layout summary. Use it before rediscovering ids manually
  • Use listWidgets() when you need the compact widget catalog. It returns: widgets (id|name|description)↓ example|Example|expanded, 4x3 widget
  • That catalog is plain text, not JSON. Do not expect widgets, items, or other object fields from it
  • If the user already named the target widget clearly, for example snake, tetris, or minesweeper, do not ask which widget. Read that widget directly by id or display name
  • On the next turn, read the visible id row directly, for example snake-game|Snake|... means use readWidget("snake-game")
  • Use that id for readWidget(), seeWidget(), patchWidget(), and reloadWidget(). Name is fallback only
  • readWidget() returns the numbered source directly in _____framework result↓
  • when widget YAML carries metadata, readWidget() includes a compact metadata: ... line before renderer↓
  • seeWidget() returns the current rendered widget HTML directly in _____framework result↓
  • seeWidget(id) strips script/style tags, inline handlers, class lists, ids, and data attrs by default
  • Use seeWidget(id, true) only when you explicitly need the full live innerHTML
  • Current Widget format after a write or reload↓ rendered↓

source↓ id: example name: Example cols: 4 rows: 3 renderer↓ 0 async (parent, currentSpace, context) => { 1 console.log("hello"); 2 }

  • Patch numbers come only from numbered renderer lines after renderer↓ inside source↓
  • Do not copy displayed line numbers into patch content
  • In prepared input, optional example turns may appear before live history, _____user = human, _____framework = runtime output, and _____transient = trailing current-space widget context

staged turns

  • listWidgets(), readWidget(), and seeWidget() are discovery calls. If the next step depends on them, end the execution there
  • If Current Space Widgets or _____framework already showed the widget id you need, skip another discovery call and move to the next step
  • After readWidget() or seeWidget(), patch on the next turn, not in the same JS block
  • After patchWidget(), renderWidget(), or reloadWidget(), use the refreshed Current Widget on the next turn if another edit is needed: rendered↓ for what mounted and source↓ for the next patch
  • Start every execution block with one short sentence saying the immediate step
  • Put that sentence on its own line. Then put _____javascript alone on the next line
  • Do not execute silently
  • Do not send only a staging sentence such as Checking widget source or Loading widget source. If you announce a widget read, list, patch, reload, or render step, the same message must execute it
  • After a successful patch or render that satisfies the request, stop and answer normally. Do not keep making more visual tweaks unless the user asked for another iteration or the runtime reported failure
  • After a successful patch or render, the next assistant turn should usually be the final user-facing answer. Do not output another promise line such as Updating... or Applying... without execution
  • Never answer with raw JS or a code fence after a widget error. Either send a proper execution message or a normal user-facing answer

examples Checking widget catalog _____javascript return await space.current.listWidgets()

Reading the widget source _____javascript return await space.current.readWidget("tetris-game")

Seeing the current rendered widget HTML _____javascript return await space.current.seeWidget("tetris-game")

After the catalog already showed snake-game, reading Snake source _____javascript return await space.current.readWidget("snake-game")

User asked for the snake widget, reading it directly _____javascript return await space.current.readWidget("snake")

bad Checking the current widget source

bad _____javascript return await space.current.readWidget("snake-game")

bad Which widget should I change?

bad const widgets = await space.current.listWidgets(); const snake = widgets.widgets.find(...)

bad return await space.current.readWidget("snake-game") return await space.current.patchWidget("snake-game", ...)

bad return await space.current.patchWidget("snake-game", ...) // success came back return await space.current.patchWidget("snake-game", ...)

bad Updating the snake widget background now Applying the color edits now

patch vs rewrite

  • Use patchWidget() for bounded edits to an existing renderer
  • Use renderWidget() for new widgets or full rewrites
  • patchWidget() is not a whole-renderer rewrite API. Do not use broad guesses like 0-999
  • Use name, cols, rows, col, row, and metadata for widget metadata changes. Use edits only for renderer lines
  • Preferred small-edit shape: [{ find, replace? }]
  • find must be one exact unique snippet copied from readWidget() output or from Current Widget source↓
  • Omit replace on a find edit to delete that snippet
  • Line-edit shape also works: [{ from, to?, content? }]
  • from and to are inclusive zero-based renderer line numbers
  • Omit to to insert before from
  • Omit content on a ranged line edit to delete
  • Common line aliases like line, startLine/endLine, range, text, and replace are tolerated, but prefer the canonical shapes above
  • Do not mix exact find edits and line edits in the same call
  • Do not overlap edits
  • The runtime applies edits from higher line numbers down to lower ones
  • If you build edits programmatically, parse the numbered renderer lines from readWidget() output or Current Widget source↓. Do not use widget.split("\n") array indexes as patch coordinates
  • If patchWidget() or renderWidget() says No files were written, the old widget file is still the source of truth. Fix and retry

example exact snippet patch _____javascript return await space.current.patchWidget("snake-game", { edits: [ { find: "parent.style.background = '#6f8f54';", replace: "parent.style.background = '#5f7f3f';" } ] })

write and reload behavior

  • readWidget() returns the numbered source readback
  • seeWidget() returns rendered HTML
  • patchWidget(), renderWidget(), and reloadWidget() return short status strings and refresh Current Widget in _____transient with both rendered↓ and source↓
  • Helpers also emit plain-text console status
  • Do not parse status strings. The execution output is enough
  • If a write or reload reports a render failure, keep fixing the widget from Current Widget source↓ before claiming success
  • Use reloadWidget(id) when you want an explicit rerun without changing source

renderer rules

  • Prefer async (parent, currentSpace, context) => { ... }
  • Render into parent
  • Use const mod = await context.import("scripts/utils.js") when several widgets need shared state, widget-to-widget communication, or other space-global functionality
  • context.import(...) resolves relative to the current space root, so copied or imported spaces keep working without hardcoded space ids, titles, or folder names
  • context.paths.scripts is the current space scripts folder
  • If several widgets need the same live shared state, import the same scripts/... module from each widget so they share one module instance on that space page
  • For very large or complex widgets, keep the renderer small and move substantial logic into scripts/*.js modules
  • If an imported module needs widget helpers or state, export functions that receive context instead of hardcoding globals or space ids
  • Do not add outer wrapper padding just to inset content. The widget shell already provides that spacing
  • Default widget card surface is #101b2d (rgba(16, 27, 45, 0.92)). Do not add another full-card background unless the content truly needs its own stage
  • Use light text and UI elements by default because widgets sit on that dark surface
  • For markdown-heavy output, use space.utils.markdown.render(text, parent)
  • If a widget should show a website, search page, URL, live web page, or browser-like surface that the agent may need to inspect or control, use <x-browser src="google.com"></x-browser> or create an x-browser element in DOM
  • Add controls="true" to x-browser only when the widget needs its own address bar and navigation controls; omit it or use controls="false" for a frameless embedded browser
  • Do not use <iframe> for general web browsing widgets. Iframes are not registered browser surfaces, so space.browser, injected browser actions, and last-interacted browser content will not work through them
  • Use an iframe only for a provider-specific embed that is intentionally not a controllable browser surface, such as a YouTube embed
  • For external HTTP reads, use fetch(url) or space.fetchExternal(url). Do not hardcode third-party CORS proxy services; the runtime already retries blocked origins through /api/proxy
  • Use space.proxy.buildUrl(url) only when you specifically need a same-origin proxied URL string for a non-fetch consumer
  • Max widget size is 24x24
  • Pick a reasonable size. Do not default to oversized cards
  • Return a cleanup function if you attach listeners, timers, or other long-lived effects
  • Do not patch unrelated page DOM
  • Do not use global plain-key listeners that interfere with chat. Require widget focus or use modified shortcuts

flow

  1. listWidgets() if you need the live catalog
  2. readWidget(id) for any existing widget you will change
  3. seeWidget(id) only when you need the live rendered DOM rather than source
  4. Next turn: patchWidget(id, ...) or renderWidget(...)
  5. Next turn if needed: use the refreshed Current Widget, then continue

Frequently asked questions

What does the Space Widgets AI skill do?

Create, patch, inspect, and remove widgets in the current space

Why use Space Widgets on TypingMind?

Because you install it once and use it with any model. Space Widgets is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Space Widgets in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/agent0ai/space-agent/tree/main/app/L0/_all/mod/_core/spaces/ext/skills/space-widgets. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Space Widgets?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Space Widgets?

As many as you like. As long as a model supports skills, you can use Space Widgets with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Space Widgets AI skill free?

Yes. It is published on GitHub by agent0ai under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇