{"version":"2.1.273","anchor":"artifact-tool-results-now-distinguish-outside-organization-w","canonical_anchor":"new-provenance-warning-for-artifacts-with-outside-contributi","heading":"Artifacts now flag content written by people outside your organization","tier":"notice","area":"Artifacts","scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.273\/e\/artifact-tool-results-now-distinguish-outside-organization-w","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.273","markdown":"### Artifacts now flag content written by people outside your organization\n\nArtifact tools and metadata now distinguish content from outside-organization writers, separate from your own co-writers\n\n**What**\n\nArtifacts (Claude Code's shareable published pages\/files) can be edited by multiple people, and previously the only distinction tracked was whether a file was \"cowritten\" or \"foreign\" to you. This release adds a more specific signal:\n\n- Artifact and asset metadata, save\/fetch responses, and tool results now carry a new `outside_writer` (also called `outsideWriter` or `outsideOrg` in some places) field, alongside the existing `cowritten`\/`foreign` fields.\n\n- When artifact tools (asset_list, asset_read, file_list, file_read, files_read, and artifact publish\/read\/db_read) return content flagged this way, the message shown now specifically says the file \"may include content from a writer outside your organization,\" instead of the more generic \"another person's page\" wording.\n\n- Artifacts containing HTML content published by an outside-organization writer now get a prompt-injection defense wrapper that tells the model to treat that content as untrusted data, not instructions, and forbids changing permissions or configuration based on it.\n\n- Telemetry now records when this outside-writer condition is detected.\n\n**Why**\n\nThis lets Claude Code warn more precisely when you're looking at content that came from someone outside your organization, rather than just a co-writer, and it hardens artifacts against prompt-injection attempts embedded in content contributed by outsiders.\n\n- Area: Artifacts\n- Tier: You'll notice\n- Useful: 2\/5\n- Signal: 2\/5"}