Lock artifacts from agent edits
Owners can lock an artifact so the agent cannot write to it. Humans editing through the UI are unaffected. A lock indicator appears next to the artifact title.
Owners decide whether the agent can write to an artifact.
Owner-only toggle
A "Lock from Agents" item in the artifact File menu blocks LLM tool-call writes against the artifact. Locked artifacts show an amber Lock icon and "Locked" pill next to the title. The toggle is gated on write access, so only owners can change lock state.
Humans keep editing
UI editing stays unaffected. The lock applies to agent-driven writes only, so a teammate can still update the artifact through the canvas, file viewer, or function editor.
Per-artifact, every renderer
The lock control mounts in the artifact view header, so every artifact renderer (canvas, attunement, file, function, app) inherits the same toggle. Lock state propagates over WebSocket to other open viewers in real time.