Artifact reads now flag more content as untrusted with a clearer warning

What

When Claude Code reads a published artifact file, it now does more work to figure out whether the content should be treated as untrusted or co-authored data, and warns about it more consistently:

• It detects whether the artifact is a recognized app type, or an HTML page containing a ws-decisions marker, and uses either signal to prepend a contentLead warning string to the returned or summarized text.
• The check for marking content as "authored by others" (untrusted or co-written) now also triggers on a new authoredByOthers field or a new internal condition, in addition to the previous checks (cowritten, typeLocked, or Slack-origin content).
• The new contentLead string is now prepended across several of the read-result templates, not just one.

Why

This makes it more likely that content written or generated outside the current session (by another tool, another app, or another author) is clearly labeled as such when Claude reads it back, reducing the chance that Claude treats untrusted or externally-authored artifact content as if it were its own prior output.