` links (data:/blob: hrefs included) and script-driven saves are inert for viewers — so never offer a file through a plain link. Artifacts render mermaid diagrams natively — markdown via ```mermaid fences, HTML via `` blocks — no external libraries involved.
**Browser storage**: `localStorage` works (so do `sessionStorage` and IndexedDB). Each artifact is served from its own origin, so what a page stores is private to that artifact, survives republishes to the same URL, and lives only in that viewer's browser — it never reaches other viewers, the viewer's other devices, or Claude. It can come back empty (a private window, cleared site data, a different browser), and in some contexts the accessor itself throws (thumbnail capture, previews, browsers set to block site data) — so wrap every read and write in try/catch and render the page correctly with no stored value. Use it for lightweight per-viewer conveniences — a remembered tab or filter, a collapsed section, an unsent draft. It is not the place for anything that must persist reliably, be shared between viewers, or be read back later by Claude.
-**Size**: The rendered page must be ${MAX_ARTIFACT_BYTES / 1024 / 1024}MB or smaller, and embedded data: URIs count toward that.
+**Size**: The rendered page must be ${MAX_ARTIFACT_BYTES/1024/1024}MB or smaller, and embedded data: URIs count toward that.
**Responsive**: Use relative units, flexbox/grid, `max-width:100%` on images. Wide content (tables, diagrams, code blocks) must scroll inside its own `overflow-x: auto` container — the page body must never scroll horizontally.
diff --git a/system-prompts/tool-description-artifact-runtime-capabilities-guidance.md b/system-prompts/tool-description-artifact-runtime-capabilities-guidance.md
index 5cbdbbb..928e700 100644
--- a/system-prompts/tool-description-artifact-runtime-capabilities-guidance.md
+++ b/system-prompts/tool-description-artifact-runtime-capabilities-guidance.md
@@ -6,4 +6,4 @@ variables:
- "ARTIFACT_CAPABILITIES_SKILL_NAME"
- "ARTIFACT_WATCH_MODE"
-->
-**Runtime capabilities** (optional): depending on what is enabled for this user, a published page can do more than static HTML — read the user's live or connected data, remember what people do on it (a poll, a sign-up sheet, a checklist, a document edited in place — the page saves new versions of itself), keep state shared across viewers, know who is viewing, ask Claude a question of its own, store files people add, or hand the viewer a file to save — declared via the `capabilities` input. **Whenever the user asks for a page that needs any of that, you MUST load the `${ARTIFACT_CAPABILITIES_SKILL_NAME}` skill BEFORE writing the artifact, and always before passing `capabilities` or writing any `window.claude.*` runtime code** — it tells you what's available to this user and how to use it. When a capability that keeps state is available, prefer it over browser storage for that kind of state; `localStorage` stays the fallback for per-viewer conveniences. Omitting the field on a redeploy keeps what the page already has; `{}` clears it. A page that saves new versions of itself ${ARTIFACT_WATCH_MODE === "none" ? "moves your local file behind it — your next publish of it then conflicts" : "reaches this session like any other republish — a republish notice on a watched artifact, or a conflict on your next publish of it — and your local file is then behind"}: re-read, merge, republish.
+**Runtime capabilities** (optional): depending on what is enabled for this user, a published page can do more than static HTML — read the user's live or connected data, remember what people do on it (a poll, a sign-up sheet, a checklist, a document edited in place — the page saves new versions of itself), keep state shared across viewers, know who is viewing, ask Claude a question of its own, store files people add, or hand the viewer a file to save — declared via the `capabilities` input. **Whenever the user asks for a page that needs any of that, you MUST load the `${ARTIFACT_CAPABILITIES_SKILL_NAME}` skill BEFORE writing the artifact, and always before passing `capabilities` or writing any `window.claude.*` runtime code** — it tells you what's available to this user and how to use it. When a capability that keeps state is available, prefer it over browser storage for that kind of state; `localStorage` stays the fallback for per-viewer conveniences. Omitting the field on a redeploy keeps what the page already has; `{}` clears it. A page that saves new versions of itself ${ARTIFACT_WATCH_MODE==="none"?"moves your local file behind it — your next publish of it then conflicts":"reaches this session like any other republish — a republish notice on a watched artifact, or a conflict on your next publish of it — and your local file is then behind"}: re-read, merge, republish.
diff --git a/system-prompts/tool-description-artifact-supporting-files-guidance.md b/system-prompts/tool-description-artifact-supporting-files-guidance.md
index c63d65a..0ba07fc 100644
--- a/system-prompts/tool-description-artifact-supporting-files-guidance.md
+++ b/system-prompts/tool-description-artifact-supporting-files-guidance.md
@@ -8,5 +8,5 @@ variables:
- "MAX_SUPPORTING_FILE_COUNT"
- "MANIFEST_TOTAL_BUDGET_BYTES"
-->
-**Supporting files**: To publish a multi-file artifact (separate CSS/JS/data/images), pass `files` as a map of published path → source file: `{"app.js": "dist/app.js", "data/points.json": "build/points.json"}`. The published path (the key) is what the HTML references (`