dsh-solo-thinking

dsh-solo-thinking addresses a common problem in brainstorming: when many directions live in one conversation, context becomes tangled, inputs cross wires, and conclusions are hard to reuse. It turns a brainstorm into an operable thinking tree, where each direction is an isolated DeepSeek Harness Session with its own state and lifecycle. Branches exchange only agent-written Handoffs, so exploration can diverge while still carrying forward explicit, reviewable context.

Its core capabilities include automatically suggesting two to four independent directions, dormant branch nodes that start only when messaged, agent-authored split handoffs, checkpoints, sibling awareness, return summaries, and replayable persisted tree state. The main input stays scoped to the current session, while the thinking tree can message selected branches directly. A full Brainstorm tab is available by default, and Better Sidebar can optionally add a right-panel view for monitoring and controlling branches.

It is best suited for users who use DeepSeek Harness for multi-option exploration, requirement breakdown, research, architecture tradeoffs, or creative divergence, and who want isolated branch contexts, automatic handoffs, and a recoverable thinking workspace without manually stitching conversations together.