use
Reach Anthropic's first-party playground capability in one step: verify the upstream playground plugin is installed, invoke its skill for interactive single-file HTML explorers (controls beside a live preview, output is a prompt to paste back), or emit the exact install commands when it is absent. Use when: 'make me a playground', 'interactive explorer for X', 'install the playground plugin', 'critique this document with approve/reject controls', or any ask to tune a parameter space visually and see the result (sliders, balancing, exploring parameters). Also carries field-tested prompt recipes, delivery guidance for cloud and remote sessions where a local browser cannot open, and known-limitation notes. This wrapper generates nothing itself; the upstream skill owns generation. Not for choosing a visual form for conversation content (a visualization capability owns that) or for throwaway variations of the project's own UI (a prototyping capability owns that).
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 30
- Forks
- 7
01 前言 Front Matter
- metadata
- { "summary": "Route playground requests to the first-party plugin, or install it, with recipes and guidance", "workflow-stage": "anytime" }
- description
- Reach Anthropic's first-party playground capability in one step: verify the upstream playground plugin is installed, invoke its skill for interactive single-file HTML explorers (controls beside a live preview, output is a prompt to paste back), or emit the exact install commands when it is absent. Use when: 'make me a playground', 'interactive explorer for X', 'install the playground plugin', 'critique this document with approve/reject controls', or any ask to tune a parameter space visually and see the result (sliders, balancing, exploring parameters). Also carries field-tested prompt recipes, delivery guidance for cloud and remote sessions where a local browser cannot open, and known-limitation notes. This wrapper generates nothing itself; the upstream skill owns generation. Not for choosing a visual form for conversation content (a visualization capability owns that) or for throwaway variations of the project's own UI (a prototyping capability owns that).
- argument-hint
- [what to explore]
- user-invocable
- true
- disable-model-invocation
- false
02 资源文件 1
- 参考文档context/consumer-notes.md1.8 KB
03 SKILL.md 正文
Use playgrounds
Purpose
Connect a playground-shaped request (an interactive parameter space the user wants to
manipulate visually, with the result flowing back as a prompt) to the maintained
first-party implementation: the playground skill from the playground plugin on the
claude-plugins-official marketplace. This skill routes, installs, and advises. It
never builds a playground itself.
Step 1: Presence check (provenance, not prose)
Determine whether the upstream plugin is installed by its identity, not its wording:
- Run
claude plugin listand look for theplaygroundplugin sourced fromclaude-plugins-official. That marketplace-qualified install record is the authoritative signal. - Where the CLI is unavailable to the session, fall back to the session's skill
listing: a bare
playgroundskill whose description says it creates interactive HTML playgrounds. Treat the description comparison as ADVISORY only; upstream may reword it at any time. Never conclude "absent" from wording alone when the name is present. - A local skill can shadow the upstream name. When the listed
playgroundskill's description clearly describes something other than an interactive-explorer builder, say a shadow is likely, then let provenance decide: with the install record confirmingplayground@claude-plugins-official, warn about the shadow and invoke the namespacedplayground:playgroundform; with no install record confirming it, take the install/guidance path, and never invoke the shadowing skill.
How a plugin skill's namespaced name and a same-named local skill resolve is Claude Code's to define; this skill always addresses the upstream skill by its namespaced form.
- Pointer: when a shadow is suspected or the bare name resolves unexpectedly, fetch Skills: Resolve skills that share a name live.
Step 2: Present, so invoke
When the upstream skill is present, invoke it via the Skill tool with the user's
brief as arguments, using the namespaced form playground:playground
(Step 1's pointer covers why). Pass the request
as stated; the upstream skill owns template choice and generation. When the user names
styles to leave out (italic accent words, pill-shaped buttons), carry that list in the
brief; when they dislike a choice in the generated page, add it to the list and invoke
again. A general "make it look less generic" only swaps one default for another. If
the invocation is refused, do not retry blind: tell the user the skill is installed but this session
could not invoke it, and suggest they run it themselves by asking for a playground in
their own words.
Step 3: Absent, so uplift
This plugin declares the upstream dependency in its manifest, so where Claude Code resolves plugin dependencies it handles the install itself. In contexts where resolution does not run (a synced or cloud session, a manual setup), emit the commands directly:
/plugin marketplace add anthropics/claude-plugins-official
/plugin install playground@claude-plugins-official
Scope guidance: user scope makes the capability available everywhere; project
scope pins it to one repository. Include the marketplace-add step only when the
install command reports the marketplace is unknown. If the install is blocked by
organization policy, say so plainly and stop; there is no workaround to offer.
- Pointer: when an install did not resolve the dependency, or reports why it could not, fetch Plugin dependencies: How dependencies behave for your users and Install plugins: Plugins with dependencies live.
Step 4: Delivery, especially off the desktop
How the upstream skill writes and opens its HTML file is that skill's to define. Feature-detect the delivery path instead of assuming its opening step reaches the user:
- A desktop session with a local browser: let the upstream flow open the file.
- A session that can publish an Artifact: offer to publish the generated HTML so the user gets a link. State that the page runs under the artifact sandbox, and confirm the copy-prompt button works there before promising the round trip.
- A session with a file-send surface (remote control, managed cloud): send the HTML file to the user to open locally.
- None of the above: report the file's absolute path and say how to open it.
Degrade visibly at every tier; never name a delivery surface as guaranteed.
- Pointer: when deciding whether the upstream flow can open the file in this session, fetch the playground skill § How to use this skill live.
Recipes (field-tested)
From the plugin author's published examples, verbatim shapes to adapt:
- "Use the playground skill to create a playground that helps me explore new layout changes to the AskUserQuestion Tool"
- "Use the playground skill to review my SKILL.md and give me inline suggestions I can approve, reject or comment"
- "Use the playground skill to tweak my intro screen to be more interesting and delightful"
- "Use the playground skill to show how this email agent codebase works and let me comment on particular nodes in the architecture to ask questions, make edits, etc"
- "Use the playground skill to help me balance the 'Inferno' hero's deck"
Repo-native recipe for this marketplace: "Use the playground skill to review plugins/<name>/skills/<skill>/SKILL.md and give me inline suggestions I can approve, reject or comment" turns skill-file review into an interactive pass.
Known limitations
Consumer guidance for the generated playgrounds lives in
context/consumer-notes.md; read it before relying on a
generated playground for anything beyond personal exploration.
Gotchas
- A listed
playgroundskill name proves presence of A skill, not THE skill, which is why Step 1 judges by install record and treats wording as advisory. - The copy-prompt button is the product; a sandboxed page (an Artifact) can restrict clipboard access, so confirm the round trip once before a long exploration.
- A summarizing fetch of the upstream marketplace can misreport its contents; verify upstream facts against raw files or the install record, never a summary.
Boundary
- Choosing the best visual FORM for conversation content is the visualization capability's job; this skill fires only when the user wants an interactive, manipulable explorer whose output returns as a prompt.
- Throwaway visual variations of your own project's UI are the prototyping capability's job; a playground explores a parameter space, not your codebase's routes.
- Generation, templates, and the playground page's own design belong to the upstream skill; nothing here overrides them.
04 同仓库更多 Skills
track
Backlog reads through the bound tracker: search open and closed items, view one item with its comments and state history. Use when another skill searches or reads tracker items.
reviewer-media
Reviewer media for one pull request: decides whether it needs a narrated walkthrough, builds the storyboard from /review:explain-change's digest, has /explainer-video:produce render it and hands the manifest to /source-control:pull-request media. Use when asked for 'reviewer media', 'a walkthrough video for this PR', 'a produced video of this change', or when pull-request ready or monitor hands a PR over.
route
Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.
route
Model and effort per agent role (orchestrator, worker, verifier, retrieval) for a multi-agent run, resolved through the bundled defaults and the user, team and personal override config layers into fenced JSON with single-agent and fan-out variants and the layer behind each value. Use when asked 'which model for the subagents', 'route the roles' or 'model routing', or before a skill passes a role's model on a spawn.
principles
Module-design vocabulary for architecture findings, and the checks built on it. Use when: 'what is a deep module', 'what is a seam', 'deep vs shallow module', 'what is the deletion test', 'leverage and locality', 'module design vocabulary', 'which word do we use for X in architecture findings'. Scanning a codebase for friction is /architecture:improve.
write
Structured five-field bug report (title, steps to reproduce, expected vs actual, justified severity, suggested fix location) from an informal description, read-only. Use when the user names a defect they observed, or asks for one written up ('report a bug', 'file a bug', 'write this up as a bug').