dsh-remote-ssh
Remote SSH execution provider for DeepSeek Harness using native DSH tools
202 results
Remote SSH execution provider for DeepSeek Harness using native DSH tools
dsh的跨会话记忆管理插件,让agent记住每一个任务的内容和进度。包含命题图、三层内存漏斗、混合向量 + 词法检索(RRF)以及浏览器端审查面板。 | Cross-session memory management for DSH — lets the agent remember the content and progress of every task while executing. Includes a proposition graph, three-layer memory funnel, hybrid vector + lexical retrieval (RRF), and a browser-side review panel.
A visual decision workbench for DeepSeek Harness: explore ideas, focus branches, turn choices into execution graphs, and run them one node at a time.
DeepSeek Harness advisor plugin: pair a fast executor with a stronger reviewer model for plan / correction / stop guidance.
DeepSeek Harness (Cordis) plugin: control Alibaba Cloud ECS instances through the local Workbench CLI — list/exec/log/upload/download/diagnose/deploy/session model tools plus a visual settings panel
运动消耗计算
Official-pattern dsh bundle: replace headless-startup with per-invocation CLI flags without writing settings
Out-of-tree DSH bundle: applies the local machine's native direnv environment to explicitly Agent/workspace-owned Bash executions and local stdio workspace MCP rows. Depends on dsh-workspace-overlay; requires a host-installed direnv; never calls allow/deny or reads .envrc itself.
Turn explicit coding-agent corrections into executable DeepSeek Harness regression tests.
DSH host bundle plugin: builds a two-level (category + skill) recall index over all local SKILL.md files and routes a natural-language intent to the best-matching skills, with three-state (hit / low-confidence / miss) fallback.
Model-facing Everything search tool (es.exe) — blazing-fast file search on Windows via the Everything engine
A public DeepSeek Harness preset for main-brain orchestration, execution gating, and reusable AI-LAW rules.
DeepSeek Harness plugin: routes every pwsh command execution on Windows through Git for Windows bash — the pwsh and bash tools actually run git bash, so the model writes bash natively.
DSH web plugin: takes over the plan review card and adds an execution-model selector at its bottom — pick which model will carry out the plan; on approve the plugin switches the session model first, then answers the review so execution runs on the chosen model.
Hardened multi-account SSH manager for DSH: profile-isolated storage, host-key pinning, Windows DPAPI secrets, pooled exec/SFTP, isolated tunnels, web terminal, cluster operations, and official agent tools.
Dual-face (host + client) plugin: independent text-output ocr_image tool + tools/execute read_image capability dispatch + ctx.ocr seam, with a conversation.input.left OCR toggle button.
Standard handoff protocol so DSH never restarts itself: a host plugin that files a JSON reset request for an external ops agent (e.g. Hermes) to execute preflight-snapshot → restart → health-check → recover, then delivers the result back to the requesting session.
Model-facing test runner tool for discovering, executing, analyzing, and reporting test results across jest/vitest/pytest/mocha/playwright.
Report dsh-TUI lifecycle state to the Herdr multiplexer (working/blocked/done in the sidebar)
DeepSeek Harness plugin: a deployment deny list of filesystem roots the model may never touch - fences read, write, search, and shell tool calls at tools/pre-execute
DSH plugin that bridges the local codebase-memory-mcp knowledge-graph server through the built-in @deepseek-ai/dsh-mcp-client, with a runtime-configurable exe path (settings namespace).
Declutter the DeepSeek Harness chat: hide agent execution-log rows (Think / context injection / tool calls), tighten message spacing and font sizes, wrap messages in rounded-corner bubbles, and inject an editable "reply concisely" instruction into the system prompt.
DSH web plugin: task execution monitor + fragmented-time micro-activity reminders (floating panel UI, activity pool editor)
Example plugin for dsh-granular-settings: three bouncing squares, one per settings scope (session, workspace, global). Switch sessions and workspaces to watch each scope behave differently. Exercises toggle, text, color, slider, number, enum, and multiselect controls.