CtrlK
BlogDocsLog inGet started
Tessl Logo

artifact-publishing

This skill should be used when the user asks to publish or share Claude Code artifacts, create a visual or interactive session output, decide whether an Artifact is appropriate, validate artifact source and output contracts, handle artifact privacy, produce share links, or choose between workspace artifacts, Claude Code Artifacts, files, and output styles.

SKILL.md
Quality
Evals
Security

Artifact Publishing

Publish, share, or validate Claude Code session output as a visual or interactive artifact. Use this skill when a user wants a live page, dashboard, walkthrough, investigation timeline, checklist, prototype, or share link instead of plain terminal text. Keep the workflow grounded in explicit source files, privacy review, and validation evidence. [DOC][CONFIG]

Treat "artifact" as two related but distinct surfaces. A Claude Code Artifact is a live, single-page output published through Claude Code and shared inside the publishing organization. A workspace artifact is a local durable file under workspace/{active}/artifacts/ or another repo-approved output path. Do not collapse them: local source files preserve traceability; Claude Code Artifacts optimize visual review and organization sharing. [DOC][CONFIG]

Supporting resources:

  • references/publishing-contract.md - Artifact definition, when to publish, source/output contract, page constraints, and update behavior. [DOC]
  • references/privacy-and-sharing.md - Privacy, sharing, retention, authorship, attachment, storage, and external-recipient boundaries. [DOC][INFERENCIA]
  • references/workspace-output-style-map.md - Decision guide for workspace artifacts, Claude Code Artifacts, ordinary files, deep links, headless runs, and output styles. [CONFIG]
  • references/validation-gates.md - Pre-publish, post-publish, link, browser, and closeout validation. [CONFIG]
  • references/official-source-map.md - Current official source URLs and drift-sensitive facts. [DOC]
  • assets/artifact-publishing-checklist.md - Reusable operational checklist. [CONFIG]
  • assets/source-output-contract-template.md - Copy-ready contract for source, output, validation, link, and privacy status. [CONFIG]
  • assets/share-decision-matrix.json - Structured routing matrix for examples and audits. [CONFIG]
  • assets/privacy-redaction-checklist.md - Privacy scan before publishing or sharing. [CONFIG]
  • scripts/check.sh - Deterministic package check for DoD terms, assets, eval cases, fixtures, and JSON validity. [CÓDIGO]

Inputs Expected

  • Publishing goal: review, executive walkthrough, PR explanation, dashboard, live checklist, data exploration, prototype, documentation page, or external handoff.
  • Audience and access: only author, named organization members, whole organization, or people outside the organization.
  • Source contract: local source path, generated HTML/Markdown path, data sources, images, CSS/JS, version, and owner.
  • Runtime gates: Claude Code surface, Team or Enterprise plan, /login authentication, Anthropic provider, organization policy, artifact setting, and disabled-environment flags.
  • Privacy posture: secrets, credentials, PII, customer data, attachments, raw logs, confidential docs, persistent storage, and audit/retention requirements.
  • Output style need: one-off published page, repeated response format, coding-preserving style, or non-coding role format.

Outputs Expected

  • A decision: publish/share a Claude Code Artifact, create a local workspace artifact, provide a file for external sharing, configure an output style, or decline publication.
  • A source/output contract naming source files, generated files, artifact link status, owner, version, validation status, and rollback/unshare path.
  • A privacy review with explicit pass/fail items and coverage_gap for unavailable gates.
  • A validation packet covering local preview, browser behavior, size/page constraints (límite vigente en references/official-source-map.md), link access, and residual risks.

Core Rules

Publish as a Claude Code Artifact only when the output benefits from visual layout, interactive controls, live status updates, or organization sharing. Keep ordinary answers, short summaries, raw logs, private notes, and external client deliverables as local files or chat output unless a published page materially improves review. [DOC][INFERENCIA]

Preserve durable source in the workspace or repo before publishing when the artifact represents work product. In foundation source, task deliverables stay in workspace/{active}/artifacts/; the published artifact is a viewing/share layer, not the only source of truth. For external recipients outside the Claude organization, produce and share the underlying HTML or Markdown file instead of a Claude Code Artifact link. [DOC][CONFIG]

Gate availability before promising a link. Claude Code Artifacts require Team or Enterprise, /login authentication, Anthropic API provider, allowed organization policy, an enabled surface, and no disabling setting or environment variable. If any gate is unverified, return coverage_gap and provide a local-file fallback. [DOC]

Build each published page as one self-contained .html, .htm, or .md file. Do not rely on external scripts, stylesheets, fonts, images, APIs, relative files, WebSockets, or backend storage. Summarize large datasets, prefer SVG/HTML/CSS over embedded raster images, and keep the rendered page below the current size limit (valor vigente en references/official-source-map.md; si no está confirmado ahí, márcalo coverage_gap). [DOC]

Run privacy review before sharing. Exclude secrets, credentials, private tokens, raw customer records, sensitive logs, unnecessary PII, and confidential attachments. State that artifact content is hosted on Anthropic-operated infrastructure and that viewers must be authenticated members of the publishing organization for Claude Code Artifacts. [DOC][INFERENCIA]

Use output styles for repeated response behavior, role, tone, or format across turns. Do not use an output style as a substitute for an artifact. If coding behavior must remain intact, set keep-coding-instructions: true; otherwise the custom style may omit Claude Code software-engineering instructions. [DOC]

Procedure

  1. Classify the surface. Decide whether the request needs a Claude Code Artifact, local workspace file, ordinary answer, output style, deep link, headless automation, channel, hook, or scheduled task. [CONFIG]
  2. Define the source/output contract. Name source paths, generated output path, expected file type, audience, share scope, validation gates, and rollback/unshare path. [CONFIG]
  3. Verify runtime gates. Check or request plan, auth route, provider, org policy, artifact setting, disabled flags, and whether the target surface can publish. [DOC]
  4. Build or request the page. Keep it single-file, self-contained, accessible, visually purposeful, and aligned with local design tokens when present. [DOC][CONFIG]
  5. Validate locally. Preview the file when possible, check console/rendering for visual outputs, scan links and text, run privacy redaction, and confirm the output still maps to the source contract. [CONFIG]
  6. Publish or share. Use the Claude Code Artifact control only after gates and privacy pass. Capture link status without exposing private URLs in public logs unless the user asks. [DOC][CONFIG]
  7. Close with evidence. Report files created, link/share status, validation results, residual coverage_gap, and how to unshare, delete, or replace the artifact. [CONFIG]

Quality Criteria

  • Surface decision separates Claude Code Artifact, local workspace artifact, ordinary file, output style, and automation alternatives.
  • Source/output contract names source path, output path, file type, owner, version, share scope, and validation gates.
  • Availability gates are verified or marked as coverage_gap.
  • Privacy review excludes secrets, credentials, PII, confidential attachments, raw logs, and unnecessary customer data.
  • Published page is self-contained, single-page, accessible, and free of blocked external dependencies.
  • Link sharing states author visibility, organization-only access, viewer sign-in requirement, and fallback for external recipients.
  • Output style guidance is used only for repeated response format, not one-off artifact publishing.
  • Closeout includes files, validation commands, residual risks, and artifact path or link status.

Usage

  • /artifact-publishing
  • Publish this Claude Code session output as an artifact
  • Should this dashboard be a workspace artifact or a Claude Code Artifact?
  • Create a source/output contract before sharing this artifact
  • Audit this artifact link for privacy and validation gaps
  • Make an output style for repeated artifact-ready summaries

Contract

  • Aceptación: artefacto/widget one-off con tema por CSS vars y fondo transparente. [EXPLICIT]
  • Límites: publica un artefacto puntual; estilos repetidos → output-style-authoring. [EXPLICIT]
  • Casos borde: artifacts = beta; degradar con gracia si no renderiza. [EXPLICIT]
  • Supuestos: mcp__visualize__show_widget disponible [SUPUESTO]; sin probe de sesión → trátalo como coverage_gap y degrada a archivo local. [SUPUESTO]
  • Trade-off: artifact único vs output-style — flexibilidad puntual vs formato reusable. [EXPLICIT]

Packet

Capas del packet, cargables bajo demanda (disciplina ICM: una capa por vez, nunca todas juntas): references/ guías de profundidad (cargar UNA por etapa) · knowledge/ cuerpo de conocimiento · prompts/ prompts listos · examples/ salida de ejemplo · agents/ subagentes del packet · templates/ plantilla de output · scripts/ automatización local · assets/ recursos estáticos.

Repository
JaviMontano/claude-plugins
Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.