v0.7.11

This release makes Goal execution more durable and evidence-backed, while making interrupted Chat sessions easier to resume and agent run details easier to inspect.

New

• Added an experimental Goal advancement loop with durable checkpoints: bounded Agent runs can save evidence, revise the current Plan, and record the next continuation, with automatic follow-up wakes for commitment and verification steps.
• Redesigned transcript command details as a terminal-style view with Shell, Task, and Markdown views, status and duration, and copyable command output.
• Added an experimental Plugins setting that keeps Hub navigation and pinned local apps opt-in.

Improved

• Improved interrupted Chat recovery: continue from the composer or press Enter to resume a preserved conversation, while a real draft or attachment still sends normally.
• Improved assignment startup recovery by checking workspace readiness and attempting dependency repair before execution.

Fixed

• Restored cht_... Chat short references across Chat commands, routes, and Agent startup context.
• Fixed Agent-created Issues entering the wrong initial state; they now start in Todo.
• Fixed Goal targets being treated as time values; they now use date-only values.

Status

• Published all public @rudderhq packages at 0.7.11 to npm; the default npx @rudderhq/cli@latest install resolves to v0.7.11.
• Tencent Cloud COS mirroring is intentionally skipped for this release.