Release Lab Agent 0.2.0 · yu7dan/lab-agent

Lab Agent 0.2.0

Add a chronological work timeline that groups continuous thinking and tool activity together, while preserving visible output/tool interleaving when the agent changes stages.
Keep long tool histories in an internal scroll area instead of expanding the whole conversation.
Render approval and follow-up prompts as bounded floating cards outside the transcript layout flow.
Allow approval and follow-up prompts to collapse into a compact pending-state control beside the workspace picker, then expand again without losing selections or custom answers.
Improve prompt density so multiple options and the custom-answer entry remain discoverable in the same panel.

Preserve manual transcript scrolling while the agent is working and avoid duplicate loading indicators.
Keep long-running work alive while progress continues, including waits for explicit approval or follow-up answers.
Continue keeping Lab Agent memory, identity, and configuration separate from Claude Code data.

The macOS build is ad-hoc signed and is not Apple-notarized. If macOS blocks the app, use System Settings → Privacy & Security → Open Anyway.