dsh-plugin-task-runner: Task Decomposition Mode
The `dsh-plugin-task-runner` plugin in DeepSeek Harness addresses the limited context windows of local large models through a “task decomposition mode.” The main agent decomposes tasks, subagents execute them in parallel within isolated, short contexts, and the main agent aggregates the results. The plugin supports concurrency control, external memory persistence, budget control, and cloud fallback, making it suitable for handling complex, long-chain tasks.
Read Moredsh-planning: DeepSeek Harness Task Planning and Decomposition Plugin
dsh-planning is a DeepSeek Harness skill plugin designed to address the challenges agents face when handling complex tasks, such as vague objectives and the difficulty of decomposing and validating tasks. The plugin provides structured task management capabilities, supporting the breakdown of vague goals into verifiable subtasks, dependency ordering, difficulty estimation, and milestone setting. It runs with the permissions of the current DSH process and is suitable for complex workflow development scenarios. For specific installation and usage instructions, refer to the source code in the GitHub repository.
Read Moredsh-adhd-copilot: Deconstruct "want to do but can't move" into a portable minimal action
`dsh-adhd-copilot` is a behavior coaching skill for the DeepSeek Harness plugin ecosystem, designed to help ADHD users reduce task initiation costs. It scans tasks and status signals via the `adhd_breakdown` tool, routing requests to specific modules based on the priority of "Emotion > Structure > Initiation," compressing complex requests into a single, executable "minimum next step action" rather than providing lengthy plans. Core mechanisms include a "Pre-flight Gate" (silent when no task signal is detected, answering trivial requests directly), an "A-G Self-Service Menu" (to assist users in choosing a direction when signals are ambiguous), and strict output discipline (first line must be an action, steps $\le$ 3, short duration per step, and prohibition of fake status updates). The plugin adopts a pure ESM architecture with zero build and zero dependencies, supporting Battery Mode and Narrative Style.
Read More