dsh-mission: Long-cycle autonomous task runtime for DeepSeek Harness

# Introduction When developing agents with DeepSeek Harness (hereinafter referred to as DSH), you will inevitably encounter a class of tasks: goals that must run for days or even longer, where progress is lost if the session disconnects or the process crashes in the middle. Even if the agent continues running and reports "finished," you cannot confirm that the task is actually completed in the environment. A single session cannot resolve these two issues. The philosophy of dsh-mission is to decompose long-term goals into a task dependency DAG and delegate them to a persistent runtime for execution: cross-session continuity, independent verification per task, replanning upon failure, and crash recovery. The model is solely responsible for proposing and planning, while state transitions are adjudicated and committed by the runtime. Below is an introduction to the positioning, core mechanisms, installation, and usage of this plugin. # What is it dsh-mission is qiao