dsh-session-handoff：长会话交接文档与新会话接力
dsh-session-handoff 是 MountBatten718 发布的 DSH 工作流插件。长会话占满上下文或体积过大时，可先确认再让 LLM 生成含目标、进度、文件与下一步的交接文档，在同一工作区创建无旧上下文的新会话并自动打开，原会话保持不变。支持手动 /handoff 与 replace、hybrid、native 等自动策略，并提供健康行、文档持久化与决策日志。安装方式为将 bundle 链入所用 profile 的 node_modules 并注册 bundles，需 Node ^22.19 或 >=24。
Read Moredsh-auto-handoff：在长会话 Step 边界交接给同工作区新会话
dsh-auto-handoff 是 qingyou002 维护的 DSH 工作流插件。它在 Step 边界（不硬中断正在运行的 Step 与工具）将长会话总结为九段式交接简报，于同一工作区创建新会话，并继承 preset、模型、权限与 Plan 模式。支持 /handoff 及 Token 阈值自动迁移，可选自动压缩。安装：dsh plugin --profile <profile> add github:qingyou002/dsh-auto-handoff，MIT 许可，基线 @deepseek-ai/dsh@0.1.5-rc.3。
Read Moreshared-handoff-dsh: Cross-Platform Task State Management and Session Handoff for DeepSeek Harness
`shared-handoff-dsh` is a DeepSeek Harness skill plugin designed to address the continuity challenges of repository-level state and session context. It encapsulates the core workflow of `shared-handoff-kit`, providing a zero-dependency, cross-platform automation solution. The plugin's core features include evidence-based session handoff and recovery, as well as repository-local task ID state management. It supports quickly starting tasks, binding sessions, or exporting context through conversational commands (such as `task-id-bootstrap` and `handoff`). Its automation mechanism is implemented via the DSH event system, requiring no manual intervention and compatible with the state layouts of Codex and Claude. The plugin natively supports Windows 10/11 differences and includes compression.
Read MoreDeepSeek Harness Plugin: dsh-handoff-button
The DeepSeek Harness plugin dsh-handoff-button provides a Handoff button after an AI response. Clicking it finalizes the current session and generates a structured document. The document uses a five-section structure covering state, objectives, progress, and more, and automatically handles file references and sensitive information redaction. Clicking the button again opens the document within the application. The plugin is installable and is restricted by a sandbox to operate only in the current session workspace, facilitating cross-session reuse or manual review.
Read MoreDeepSeek Harness Plugin: dsh-handoff - Deterministic Markdown Export for Session Handoff
`dsh-handoff` is a DSH plugin that automatically generates deterministic Markdown handoff documents during multi-turn sessions. It does not call an LLM; instead, it organizes goals, completed or pending tasks, key files, and statistics based on the session event stream. The plugin includes path-fencing security constraints and fallback strategies, making it suitable for quickly handing over workflow progress to a successor at session end or during context switching.
Read Moredsh-session-handoff: DeepSeek Harness Session Handoff Observation and Trigger Plugin
`dsh-session-handoff` is a host plugin for DeepSeek Harness (DSH) used to automatically determine when to perform a session handoff during long conversations or complex tasks. It adopts an “observe-decide-trigger” architecture, recording session metrics (such as turns and steps) and comparing them against thresholds in a policy file, then appending deterministic recommendations to the tail when the session state becomes anomalous. The plugin strictly follows read-only policies, append-only injection discipline, and approval gating mechanisms, ensuring that handoff logic is decoupled from observation logic and ultimately invoking the CLI to execute operations.
Read Moredsh-handoff: A Session Handoff Documentation Plugin for Agent Session Compression and Relay
dsh-handoff is a skill plugin for DeepSeek Harness (DSH) designed to address the issues of context congestion and information loss caused by an increasing number of conversation turns. It generates a “session handoff document” to compress and organize complex conversational context, enabling standardized session handoffs. The plugin supports context compression and rephrasing, making it suitable for scenarios involving multi-agent collaboration or periodically restarting sessions while retaining key information. Users should refer to GitHub for installation instructions.
Read More