Mark v1.9.5-v1.10.15 as published

This commit is contained in:
tux tucker 2026-05-18 22:20:08 -04:00
parent 64bc4dc07f
commit a384177b95

View file

@ -4,6 +4,8 @@
- Fixed artifact extraction silently dropping all artifacts in conversations where Anthropic's `enabled_artifacts_attachments` setting is false. In that mode Claude uses the skills-runner `create_file` MCP tool instead of the legacy `artifacts` tool — same `display_content` shape (json_block with language / code / filename), different `tool_use.name`. The v1.9.1 strict allowlist (`name === 'artifacts'`) was rejecting it. Allowlist now accepts `'artifacts'` or `'create_file'`. Added two regression tests covering the create_file pattern and a negative case for non-artifact skills tools (`view`, etc.) that share the same display shape.
_Published_
## [1.10.14]
- Single-conversation export toast now includes the artifact count when applicable: `Exported: X with N artifact(s)` when artifacts were extracted, `Exported: X` otherwise. Tracked via a function-scope `artifactCount` set inside the extraction branch so the unified post-save toast can read it without restoring the old double-toast pattern.